A TSN-based industrial real-time data transmission guarantee method and system

Abstract

The embodiments of the present disclosure provide a method and system for ensuring industrial real-time data transmission based on TSN. Applied to the field of industrial communications and intelligent scheduling, the method includes: obtaining industrial field network node information and completing TSN topology deployment and time synchronization, extracting communication characteristics and historical load data of various task flows, constructing a unified time series and inputting it into the prediction model to generate load prediction results. Combined with task parameters, scheduling strategy lists and prediction results, the TSN digital twin network model is input to carry out simulation, evaluate performance indicators under multiple strategies, select the optimal scheduling strategy and send it to the TSN switching equipment, and realize accurate scheduling control configuration of various task flows. This solution realizes the intelligent evaluation and automatic optimization and distribution of scheduling strategies, which not only significantly improves the real-time and deterministic communication capabilities of industrial networks, but also enhances resource utilization efficiency and the adaptability of scheduling decisions, ensuring that critical task flows are transmitted on time and reliably.

Description

Technical Field

[0001] The present disclosure relates to the technical field of industrial communication and intelligent scheduling, and in particular to a method and system for ensuring industrial real-time data transmission based on TSN. Background Art

[0002] Traditional industrial networks, limited by the inherent non-deterministic nature of Ethernet, struggle to meet the stringent microsecond latency and nanosecond clock synchronization requirements of scenarios such as high-precision motion control and multi-device collaborative operations. The emergence of Time-Sensitive Networking (TSN) technology provides a key path to addressing this challenge. Through mechanisms such as time synchronization, traffic scheduling, and resource reservation, it enables deterministic transmission of multiple service flows on a unified Ethernet physical layer. However, given the complex and ever-changing industrial field environments and the high heterogeneity of equipment, traditional TSN deployments still face challenges such as insufficient dynamic adaptation capabilities and difficulty ensuring QoS for multiple services. Building an adaptive TSN network architecture to achieve efficient and reliable transmission of real-time industrial data has become a core bottleneck hindering the development of intelligent manufacturing.

[0003] Existing industrial network deployment solutions mostly use statically configured TSN topologies, with the gated control list (GCL) and traffic shaping parameters of each network node preset through a centralized management platform. In terms of time synchronization, they mainly rely on the IEEE1588v2 protocol to achieve clock alignment between nodes, and cooperate with boundary clocks or transparent clock mechanisms to reduce synchronization errors. For service flow scheduling, static scheduling strategies based on credit shaping (CBS) or asynchronous traffic shaping (ATS) are generally adopted, combined with reserved bandwidth and priority queues to achieve differentiated services. In terms of dynamic adaptability, some solutions introduce simple load monitoring mechanisms and perform periodic parameter adjustments based on historical data statistics, but lack in-depth coupling analysis of service characteristics and network status.

[0004] The static configuration mode of existing technology cannot adapt to the dynamic changes of industrial sites. When equipment is added or reduced, task flows are changed, or network loads fluctuate, manual intervention and reconfiguration are required, resulting in long network adjustment cycles and high operation and maintenance costs. At the same time, traditional scheduling strategies only consider single-dimensional delay or bandwidth indicators, and fail to establish multi-parameter joint optimization models such as periodicity, fault tolerance, and priority. Resource competition or service degradation are prone to occur in complex industrial scenarios. In addition, the accuracy of time synchronization is limited by the hardware clock source and transmission medium. In cross-domain networking scenarios, the error accumulation effect is obvious, making it difficult to meet the nanosecond-level synchronization accuracy requirements of multi-workshop collaborative control. Summary of the Invention

[0005] In order to address the deficiencies of the existing technology, the present disclosure provides a method and system for industrial real-time data transmission based on TSN. The present disclosure solves the problem that the static configuration mode of the existing technology cannot adapt to the dynamic changes of the industrial site. When equipment is added or reduced, the task flow changes, or the network load fluctuates, manual intervention and reconfiguration are required, resulting in a long network adjustment cycle and high operation and maintenance costs. At the same time, the traditional scheduling strategy only considers the single-dimensional delay or bandwidth indicators, and fails to establish a multi-parameter joint optimization model such as periodicity, fault tolerance, and priority. In complex industrial scenarios, resource competition or service degradation is prone to occur. In addition, the time synchronization accuracy is limited by the hardware clock source and the transmission medium. In the cross-domain networking scenario, the error accumulation effect is obvious, and it is difficult to meet the requirements of multi-workshop collaborative control for nanosecond-level synchronization accuracy.

[0006] According to a first aspect of the present disclosure, a method for ensuring industrial real-time data transmission based on TSN is provided, comprising:

[0007] Obtain network node information at the industrial site, perform TSN topology deployment and configuration based on the network node information, and obtain an industrial network topology structure that supports the TSN protocol;

[0008] Obtain the local system time information of each TSN network node in the industrial network topology, accurately synchronize the local system time information, and obtain a unified system time benchmark for the entire network;

[0009] Acquiring historical communication data and service characteristic data of various task flows in the industrial site, performing feature extraction on the historical communication data and service characteristic data, and obtaining task parameters of the various task flows; wherein the task parameters include periodicity, maximum delay tolerance, bandwidth requirement, jitter tolerance, and priority level;

[0010] Obtain historical load data of various task flows, time-align the historical load data according to the system time base, construct a unified time series, input the unified time series into a preset prediction model, and obtain load prediction results for various task flows;

[0011] Input the system time base, task parameters, preset scheduling strategy list, and load forecast results into the preset TSN digital twin network model, perform simulations of multiple scheduling strategies, and obtain performance indicator data under each scheduling strategy;

[0012] An optimal scheduling strategy is determined based on the performance indicator data and the task parameters, and the optimal scheduling strategy is sent to the TSN switching device to complete the scheduling control configuration of various task flows in the network.

[0013] According to a second aspect of the present disclosure, a TSN-based industrial real-time data transmission assurance system is provided, which is configured to execute the method according to the first aspect, including:

[0014] The TSN network topology deployment module is used to obtain network node information of the industrial site, perform TSN topology deployment and configuration according to the network node information, and obtain an industrial network topology structure that supports the TSN protocol;

[0015] The system time synchronization module is used to obtain the local system time information of each TSN network node in the industrial network topology, accurately synchronize the local system time information, and obtain a unified system time benchmark for the entire network;

[0016] A task parameter extraction module is used to obtain historical communication data and service characteristic data of various task flows in the industrial field, perform feature extraction on the historical communication data and service characteristic data, and obtain task parameters of the various task flows; wherein the task parameters include periodicity, maximum delay tolerance, bandwidth requirement, jitter tolerance, and priority level;

[0017] A load prediction module is used to obtain historical load data of various task flows, time-align the historical load data according to the system time reference, construct a unified time series, input the unified time series into a preset prediction model, and obtain load prediction results for various task flows;

[0018] The TSN simulation module is used to input the system time base, task parameters, preset scheduling strategy list and load forecast results into the preset TSN digital twin network model, perform simulations of multiple scheduling strategies, and obtain performance indicator data under each scheduling strategy;

[0019] The scheduling optimization module is used to determine the optimal scheduling strategy based on the performance indicator data and the task parameters, send the optimal scheduling strategy to the TSN switching device, and complete the scheduling control configuration of various task flows in the network.

[0020] According to a third aspect of the present disclosure, an electronic device is provided. The electronic device includes: a memory and a processor. The memory stores a computer program. When the processor executes the program, the method described above is implemented.

[0021] In the TSN-based industrial real-time data transmission guarantee method and system provided above, the disclosed embodiment realizes the intelligent evaluation and automatic optimization and distribution of scheduling strategies by integrating system time synchronization, task parameter modeling, load prediction and TSN digital twin simulation, which not only significantly improves the real-time and deterministic communication capabilities of the industrial network, but also enhances resource utilization efficiency and the adaptability of scheduling decisions, ensuring the reliable and on-time transmission of critical task flows. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the embodiments of the present disclosure or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present disclosure. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0023] Figure 1 A schematic diagram of a method for ensuring industrial real-time data transmission based on TSN according to an embodiment of the present disclosure is shown;

[0024] Figure 2 A schematic diagram of a method for ensuring industrial real-time data transmission based on TSN according to an embodiment of the present disclosure is shown;

[0025] Figure 3 A schematic block diagram of an industrial real-time data transmission assurance system based on TSN according to an embodiment of the present disclosure is shown;

[0026] Figure 4 A block diagram of an exemplary electronic device according to an embodiment of the present disclosure is shown. DETAILED DESCRIPTION

[0027] Various exemplary embodiments of the present disclosure will now be described in detail with reference to the accompanying drawings. It should be noted that unless otherwise specifically stated, the relative arrangement of components and steps, numerical expressions and numerical values ​​set forth in these embodiments do not limit the scope of the present disclosure.

[0028] Those skilled in the art will understand that the terms "first", "second" and the like in the embodiments of the present disclosure are only used to distinguish different steps, devices or modules, etc., and do not represent any specific technical meaning, nor do they represent the necessary logical order between them. It should also be understood that in the embodiments of the present disclosure, "multiple" may refer to two or more, and "at least one" may refer to one, two or more. It should also be understood that any component, data or structure mentioned in the embodiments of the present disclosure can generally be understood as one or more, unless explicitly defined or given a contrary revelation in the context. In addition, the term "and / or" in the present disclosure is merely a description of the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in the present disclosure generally indicates that the associated objects before and after are in an "or" relationship. It should also be understood that the description of each embodiment in the present disclosure emphasizes the differences between the embodiments, and the same or similar aspects thereof can be referenced to each other. For the sake of brevity, they will not be described one by one.

[0029] At the same time, it should be understood that for ease of description, the dimensions of the various parts shown in the drawings are not drawn according to the actual proportional relationship. The following description of at least one exemplary embodiment is actually only illustrative and is in no way intended to limit the present disclosure and its application or use. Technologies, methods and equipment known to those of ordinary skill in the relevant art may not be discussed in detail, but where appropriate, the technologies, methods and equipment should be considered part of the specification. It should be noted that similar numbers and letters represent similar items in the following figures, so once an item is defined in one figure, it does not need to be further discussed in subsequent figures.

[0030] To make the objectives, technical solutions, and advantages of the embodiments of the present disclosure more clear, the technical solutions in the embodiments of the present disclosure will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present disclosure, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present disclosure without making any creative efforts shall fall within the scope of protection of the present disclosure.

[0031] Figure 1 This is a flow chart of a TSN-based industrial real-time data transmission assurance method provided in an embodiment of the present disclosure. The method of the embodiment of the present disclosure aims to achieve accurate detection of large and small objects in images.

[0032] S101, obtaining network node information of an industrial site, performing TSN topology deployment and configuration according to the network node information, and obtaining an industrial network topology structure supporting the TSN protocol.

[0033] The industrial site can refer to the physical environment where the industrial control system (ICS) is actually deployed, including the field control network composed of sensors, actuators, PLCs (programmable logic controllers), industrial robots, edge computing devices, data acquisition equipment, control terminals, etc.

[0034] Network node information can refer to the identification and attribute data of each communication participating unit in the industrial network, including node ID or MAC address; the port and topological location to which the node is connected; the supported communication protocol stack (such as whether TSN is supported); node type (such as controller node, sensor node, actuator node, switching node); network interface parameters (IP address, bandwidth capacity, delay characteristics); and link status information with other nodes.

[0035] The TSN protocol is a set of Ethernet real-time transmission extension standards developed by the IEEE802.1TSN working group to support industrial-grade deterministic communications. Its key capabilities include time synchronization (IEEE802.1AS): ensuring that all network devices share a unified clock; time-aware scheduling (IEEE802.1Qbv): arranging data transmission windows according to a periodic table; traffic shaping and queuing (such as IEEE802.1Qav and Qch); path redundancy and seamless switching (such as 802.1CB); and flow identification and management mechanisms (such as 802.1Qci).

[0036] The industrial network topology structure can refer to a communication network structure diagram that supports the TSN protocol, which is constructed based on the physical connection relationship and logical transmission strategy of the industrial site, including the physical and logical connection relationship between nodes (chain, star, ring, hybrid, etc.); TSN switching equipment deployment location and connection strategy; path planning and redundancy design (for fault tolerance); mapping path of communication task flow; bandwidth, latency, synchronization characteristics and other parameters of each link.

[0037] Through network scanning, device reporting, and configuration parsing, detailed information about each network node in the industrial field is obtained. This information includes the node's unique identifier (such as MAC address, device serial number), network interface parameters (such as IP address, number of ports, supported bandwidth, and protocol type), device type (such as TSN switch, PLC, sensor, edge gateway, etc.), and the physical connection relationship between nodes. Industrial Ethernet management protocols (such as SNMP, LLDP, OPC UA) can be used to communicate with field devices to collect and parse the link information and capability descriptions provided by the devices. The collected network node information is used to identify and filter key network devices with TSN capabilities, such as TSN switches, TSN-capable terminal devices, and edge controllers, and mark them as TSN nodes. Based on the functional support information of these TSN nodes, their functions are verified and their roles are divided, such as specifying the master clock node (GrandmasterClock) in IEEE802.1AS time synchronization, configuring boundary clocks, and slave nodes.

[0038] Subsequently, the TSN topology configuration process is executed, which specifically includes the following operations: Time synchronization configuration: Based on the clock support capabilities and connection relationships of each network node, according to the IEEE802.1AS protocol, precise time protocol parameters are configured for all TSN nodes, synchronization message forwarding paths are established, and the clocks of all nodes in the network are aligned to a unified system time reference; Path and resource configuration: Based on the forwarding capabilities, resource status, and link topology information of the network nodes, according to the IEEE802.1Qcc standard, a centralized network configuration mechanism is used to plan the task flow transmission path and configure the forwarding rules, queue resources, and traffic scheduling strategy in the switching nodes for each path; Scheduling strategy distribution: Based on the network node's support for scheduling protocols (such as IEEE802.1Qbv), a GCL (Gate Control List) containing scheduling periods and sending time slots is generated and distributed to each switching node to achieve precise timing control of task flow transmission. TSN function activation and verification: Based on the known network node distribution, functional characteristics, and connection relationships, TSN function activation and connectivity verification are performed on each node to ensure that the scheduling table is loaded successfully, the time synchronization status is stable, and the path planning is correct. Through the above steps, we successfully completed the collection and screening of industrial field network node information, realized the deployment and configuration of the network topology based on the TSN protocol, and finally obtained the industrial network topology structure that supports the TSN protocol, providing a basic guarantee for the determinism of industrial real-time data transmission.

[0039] S102, obtaining local system time information of each TSN network node in the industrial network topology, accurately synchronizing the local system time information, and obtaining a unified system time reference for the entire network.

[0040] TSN network nodes refer to key nodes selected from all network nodes that possess TSN capabilities and play a role in TSN deployment. These include TSN switches; terminal devices that support the TSN protocol stack; time synchronization master and slave nodes; and devices involved in scheduling, configuration, and task forwarding. They represent a subset of "industrial field network node information" resulting from functional identification, protocol capability filtering, and task adaptation.

[0041] Local system time information refers to the current time value of the system clock maintained within each TSN network node, representing that node's understanding of the "present moment." Due to minor differences in each node's clock hardware, these local times may deviate. This information may include the current local timestamp; clock frequency and drift; time synchronization protocol support information (such as synchronization status and slave / master clock role); and data such as delay and offset of time synchronization messages (PTP, IEEE1588, or 802.1AS).

[0042] The system time base can refer to a unified standard time reference established across all nodes in the network through the TSN time synchronization protocol (such as IEEE802.1AS). It ensures that the local clocks of all TSN network nodes remain highly consistent, allowing cross-device scheduling and control operations to be performed based on a unified timeline.

[0043] After completing the TSN industrial network topology deployment and identifying TSN protocol-supported nodes, in order to achieve time synchronization, the entire network time synchronization process is first started based on IEEE802.1AS (General Precision Time Protocol, gPTP). Specifically, by enabling the gPTP protocol stack on each TSN network node, the local system time information of the node is read and extracted, that is, the current timing value of each internal high-precision clock, which is usually provided by the hardware timestamp unit integrated in the device or the system counter driven by the crystal oscillator. Subsequently, according to the best master clock election algorithm defined by IEEE802.1AS, the clock accuracy level, stability parameters, priority configuration and other factors of each TSN node are compared, and the node with the best performance is automatically selected as the master clock. The node selected as the master clock will periodically broadcast Sync messages and Follow_Up messages to spread its system time information to the entire network.

[0044] When other nodes receive these synchronization messages, they use their local timestamp capture mechanism to record the time of receipt and, using the PeerDelay measurement mechanism in IEEE802.1AS (using the Pdelay_Req and Pdelay_Resp messages), estimate the link transmission delay between themselves and the master clock. They calculate clock offset based on the master clock time carried in the Sync and Follow_Up messages, the local reception time, and the link delay. Once this is complete, they align their system time with the master clock by adjusting the local timer frequency or stepping the current time.

[0045] In scenarios with complex topologies or multiple subnets, some TSN switching nodes can be configured as boundary clocks. These nodes serve as both time slave nodes for a certain link and time master nodes for another link, thereby achieving time coordination across domain synchronization links and avoiding error accumulation caused by clock hierarchical transmission.

[0046] To ensure time synchronization accuracy meets TSN requirements (typically ±1 microsecond or finer), the system sets a synchronization message broadcast period (e.g., every 125ms or less). During the synchronization process, sliding average, synchronization jitter filtering, and deviation monitoring mechanisms are enabled to dynamically assess synchronization accuracy. When the time deviation between all nodes is less than a preset threshold and synchronization remains stable for a certain period, the establishment of a unified system time base for the entire network is confirmed.

[0047] This time base will serve as the key foundation for subsequent timing-sensitive operations such as scheduling simulation, schedule table generation (such as GCL), and task sending control, ensuring that all types of real-time tasks in the entire industrial control network have a unified and high-precision time coordinate system.

[0048] S103, obtaining historical communication data and service characteristic data of various task flows in the industrial site, performing feature extraction on the historical communication data and service characteristic data, and obtaining task parameters of the various task flows; wherein the task parameters include periodicity, maximum delay tolerance, bandwidth requirement, jitter tolerance, and priority level.

[0049] Task flows refer to data communication flows generated by different business systems or devices in industrial sites, with specific transmission characteristics and real-time requirements. These can include control instruction flows (such as PLC control signals to actuators), condition monitoring flows (such as data uploaded by temperature, pressure, and position sensors), alarm information flows (such as alarm data triggered by fault detection), video surveillance flows (such as real-time HD video transmission), and system configuration flows (such as system parameter updates and remote configuration information). Each type of task flow typically corresponds to a specific business scenario and has different quality of service requirements.

[0050] Historical communication data can refer to the communication records of the task flow collected during the network operation over a period of time. It can include link quality indicators such as message arrival timestamp, message size (number of bytes), message interval (period), actual delay and jitter data, packet loss rate, and number of retransmissions.

[0051] Business characteristic data can refer to information related to the business background, importance and operation logic of the task flow, such as task type (periodic or event-triggered), device role (master / slave / sensor / actuator), data generation mechanism (timing, edge triggering, etc.), and the level of demand for real-time and reliability.

[0052] Task parameters are a set of key indicators describing task flow scheduling requirements, generated by extracting features from historical communication data and service characteristics. They are primarily used for TSN scheduling simulation and decision optimization. These indicators include periodicity, which indicates whether the task flow is periodic, and its fixed or variable transmission period, such as 10ms or 100ms. The shorter the period, the higher the real-time requirements.

[0053] Maximum delay tolerance: The maximum transmission delay allowed for a task flow from the source node to the destination node. For example, for a servo control signal, this value may be less than 1ms.

[0054] Bandwidth requirement: This refers to the amount of data that a task flow needs to transmit per unit time, usually expressed in Mbps. For example, a video stream may require more than 5Mbps, while sensor data may only require tens of Kbps.

[0055] Jitter tolerance: This measures the stability requirements of the arrival interval of a task flow and refers to the maximum acceptable transmission time fluctuation range, such as ±20μs.

[0056] Priority level: used to describe the priority scheduling order of task flows during scheduling. It is usually mapped to TC (Traffic Class) based on IEEE802.1Q priority, such as real-time control flow: highest priority, alarm flow: higher priority, normal status reporting: medium priority, file update or non-real-time data: lowest priority.

[0057] By deploying industrial communication data acquisition units (such as industrial firewalls, mirror port listeners, industrial protocol gateways, etc.), deep packet capture and protocol identification can be performed on data packets on key network links. This process uses parsing modules that support commonly used industrial field protocols (such as Profinet, EtherCAT, and OPCUA) to perform protocol layer decoding and task flow identification on the network messages collected in real time, and classify communication behaviors in the form of data source-data destination-communication content triples to obtain communication flow identifiers with business semantics. For each type of identified task flow, its historical communication data is recorded, including message sending / receiving timestamps (for calculating intervals and delays), message size and total number of bytes (for calculating bandwidth), actual path delay and hop count (for network performance evaluation), whether retransmissions are performed, whether congestion and packet loss occur, and other communication status information. These data are stored in local edge nodes or industrial cloud platforms to form a historical communication log database with time continuity.

[0058] In parallel, business characteristic data is collected from the equipment side. By reading PLC / industrial controller configuration files, task schedules, and OPCUA server node structures, business metadata such as the trigger mechanism (cycle / event), control logic priority, real-time requirements, and fault level for each task type can be obtained. Furthermore, industrial system engineering design drawings (such as BOMs and IO lists) and BIM integration systems can also provide data sources at the business layer.

[0059] Next, the task parameter extraction algorithm is run on the edge computing node or data analysis platform. Periodic parameter extraction begins by analyzing the timestamp sequence of the task flow using a sliding window algorithm. If the standard deviation of the intervals between adjacent messages is less than a preset threshold, the task is identified as periodic, and the average interval is used as its period value. If there is no stable interval, the task is considered an event-triggered task. Maximum delay tolerance can be extracted based on the control response time requirements specified in the service description document, or by estimating the maximum delay value before an anomaly in historical communication data. Correction is then made based on the fault tolerance of the service system. Bandwidth requirements are calculated by statistically analyzing the total amount of data transmitted per unit time by the task flow to form two metrics: average bandwidth and peak bandwidth. These metrics are used to guide TSN resource allocation. Jitter tolerance is calculated by statistically analyzing the deviation between the time interval of a periodic task and the theoretical period to determine the maximum allowable jitter range, reflecting its dependence on time stability. Priority level is determined based on a comprehensive consideration of the security level, control level, and alarm level of the device to which the task belongs. It can also reference the PCP field in the VLAN tag, industrial QoS policy configuration, or priority rule mapping manually set by system engineers.

[0060] Ultimately, all extracted task parameters are encapsulated into a unified description format, such as a structured table or standardized JSON object, and associated with the task flow's identification information to form a task parameter library. This task parameter library can be directly called by the subsequent TSN digital twin simulation platform or network configuration module to perform operations such as network scheduling simulation, path planning, bandwidth allocation, and priority mapping, ensuring TSN real-time performance tailored to business needs.

[0061] S104, obtaining historical load data of various task flows, time-aligning the historical load data according to the system time reference, constructing a unified time series, inputting the unified time series into a preset prediction model, and obtaining load prediction results of various task flows.

[0062] Historical load data refers to network load characteristics such as bandwidth usage, data frame transmission frequency, and peak data traffic generated by various task flows in the past period of time. This includes the total number of bytes transmitted by a task flow in each time slice (e.g., 1 second, 100 milliseconds); the statistical distribution of the number and size of packets; peak and average bandwidth; the time density of packet arrivals (i.e., load bursts within a short period of time); changes in network link utilization; forwarding queue length or latency indicators along each hop; and whether the task experiences packet loss or retransmissions.

[0063] A unified time series refers to a standardized timeline data structure formed by aligning and resampling the historical load data collected by each network node based on the unified system time reference obtained in the previous step. Because different devices may collect data at different times or out of sync, a unified time reference (such as the synchronization time provided by IEEE802.1AS) can be used to align the data from all nodes in a strict chronological order, constructing a data structure of the following form:

[0064] Time points: t0, t1, t2, ... (fixed intervals, such as every 10ms)

[0065] The load value of each task flow corresponding to each time point: for example, task A sends 128 bytes at time t0, task B sends 512 bytes, and so on.

[0066] Pre-set prediction models can refer to machine learning or statistical models for load trend forecasting that are pre-trained or dynamically loaded in the data analysis platform or edge node. Common model types include time series models, such as ARIMA and Prophet, which are used to model and forecast periodic and trending loads; neural network models, such as recurrent neural networks like LSTM and GRU, which are suitable for forecasting task flow loads with long-term dependencies; time convolutional networks (TCNs), which are suitable for capturing local bursts and multi-scale patterns in loads; hybrid models, which combine periodic rules and data-driven algorithms, and are suitable for industrial flows with partially deterministic business characteristics; and lightweight edge models, such as approximate prediction methods based on sliding window weighted averages, which are suitable for resource-constrained edge devices.

[0067] Load forecasting results refer to the output of a pre-set prediction model after inputting a unified time series into a specific time series. These results reflect the load trends of various task flows over a specific timeframe (e.g., the next 10 seconds). These forecasts include the estimated bandwidth requirements of each task flow at specific points in the future (e.g., the predicted number of bytes sent per second over the next 5 seconds); possible bandwidth peak times (predicted traffic spikes); the presence of bursty loads (e.g., out-of-cycle control commands and alarm data); and time windows for load overlap or conflict between task flows. Predicted communication cycle: This refers to the expected triggering period for the task flow over a specific timeframe, measured in milliseconds. Predicted bandwidth requirement: This indicates the estimated bandwidth usage of the task flow per unit time, measured in Mbps or kB / s. Predicted maximum latency: This refers to the predicted maximum end-to-end transmission delay after network scheduling, measured in milliseconds. Predicted jitter range: This refers to the fluctuation range of latency, reflecting scheduling stability. Predicted packet loss rate: For task flows with high real-time reliability (e.g., remote control signals), the packet loss rate must be extremely low.

[0068] To obtain historical load data for various task flows during industrial network operation, edge traffic sensing and analysis technologies are deployed within a TSN-enabled network architecture, combined with the port traffic collection capabilities of TSN switching devices, to continuously monitor identified task flows. Based on the previously identified communication triplet (data source, data destination, and communication content), this process continuously collects load-related metrics such as the number of messages, total message size, send time, and receive time for each task within the network within each time period. To ensure a consistent time reference for multi-source data, the industrial network utilizes IEEE 1588 precise time synchronization technology to establish a time synchronization system based on the system time base, mapping the load data of all task flows to a unified network-wide time axis. Specifically, the raw data from each load collection point is timestamped and aligned based on the system time base, constructing a standard time slice sequence with a fixed time granularity (e.g., 10 milliseconds). Within each time slice, statistics are collected for the corresponding task flow's instantaneous bandwidth usage, total number of transmitted bytes, or communication density, thereby generating a structured, unified time series. This time series is organized according to the task flow dimension and forms a multi-dimensional time series data structure as the input for the next prediction. Subsequently, the unified time series is input into the pre-trained load prediction model, which is based on the behavioral characteristics of industrial loads and is constructed in combination with technologies such as long short-term memory neural networks (LSTM), gated recurrent units (GRU), or time series regression analysis. The model uses historical bandwidth occupancy patterns, fluctuation patterns, communication cycles, and other time series features to output load prediction results within the prediction period, that is, the expected bandwidth requirements, peak communication loads, and load trend changes of various task flows at different time points in the future time window. The prediction results serve as key inputs to participate in the TSN digital twin simulation and scheduling optimization process, improving the adaptability of industrial networks to the dynamic changes of real-time task flows and the accuracy of resource scheduling.

[0069] The training process of the preset prediction model is:

[0070] The training process of the preset prediction model mainly relies on a large amount of historical load data collected from industrial sites as training samples, and combines it with a unified system time base for data alignment processing to construct a standardized unified time series data set as the input feature of the model. First, during the data collection phase, it is necessary to continuously monitor the network communication behavior of various task flows in a fine-grained manner and extract the key load indicators of each task flow at a unit time granularity (such as every 10 milliseconds, every 100 milliseconds), including the number of messages, the total number of message bytes, the instantaneous bandwidth occupancy, the peak rate, the average interval time, the number of congestion occurrences, etc. All collected data must be timestamp aligned using a unified high-precision time synchronization mechanism of the industrial network (such as IEEE1588PTP) to ensure that data from different collection points are under the same time base, forming a time-continuous and ordered load time series data set.

[0071] During the sample construction phase, the system performs sliding window processing on the time series data of each communication flow at the task flow granularity, setting a fixed window length (such as 30 time steps) as the model input sequence. Each time step contains multiple feature dimensions at that time point (such as bandwidth, number of packets, average latency, etc.), thus forming the model's input matrix. The corresponding actual load results are key load indicators within several time steps after the prediction window (such as instantaneous bandwidth occupancy or total load for the next five time steps) to achieve prediction of future load trends. Therefore, the actual load results constitute the prediction target within this time period and can be set as a single-point prediction (predicting the next time point) or a sequence prediction (predicting a sequence of several future time points) to guide the model's supervised training process.

[0072] During the model design and training phase, considering the significant time dependence and cyclical fluctuations of industrial task flow load data, recurrent neural network structures from deep learning, specifically long short-term memory networks (LSTMs) or gated recurrent units (GRUs), were selected to model the input load time series. This type of structure can effectively capture potential cyclical and sudden changes in historical sequences and has the ability to jointly predict short-term and medium- to long-term trends. During training, the mean squared error (MSE) or mean absolute error (MAE) loss function is used to calculate the difference between the model output and the corresponding actual load results, guiding the gradient update of the network parameters and continuously improving prediction accuracy.

[0073] To enhance the model's generalization and stability, the training dataset should cover diverse network operating scenarios, such as normal communication, mild congestion, link jitter, and sudden high-load conditions. All feature data should be normalized to eliminate the impact of dimensionality differences on training performance. Furthermore, in the post-training phase, early stopping mechanisms, overfitting regularization terms, and cross-validation can be used to optimize the model's hyperparameter combination to ensure the model's adaptability in real-world deployments.

[0074] Ultimately, the trained prediction model is capable of rapidly forecasting the changing trends of various task flow load indicators over short periods of time based on a unified time series input. The predicted output is an approximation of actual future load results, such as bandwidth requirements and transmission load size for the next five seconds. Once deployed, the model can run in real time on edge computing nodes or industrial cloud platforms, providing accurate and forward-looking load forecasts for subsequent TSN scheduling simulations and policy decisions. Furthermore, the system can regularly incorporate newly collected data for fine-tuning and retraining, continuously adapting to evolving task flows, business changes, and network status changes, maintaining high-quality prediction performance and scheduling decision-making support.

[0075] S105: Input the system time benchmark, task parameters, preset scheduling strategy list and load forecast results into the preset TSN digital twin network model, perform simulations of multiple scheduling strategies, and obtain performance indicator data under each scheduling strategy.

[0076] The pre-set TSN digital twin network model can be a simulation system or digital replica that simulates the operational behavior of a real-world TSN industrial network, including communication flow scheduling, forwarding, latency, packet loss, and other behaviors. This model has the capability of a "virtual network runtime environment."

[0077] The preset scheduling strategy list can be a strategy set, which pre-lists the scheduling method combinations that can be used for TSN network simulation and optimization. Each scheduling strategy corresponds to a set of scheduling algorithms and parameter configurations, and the simulation system will simulate and execute each strategy in turn.

[0078] Scheduling policy refers to a set of rules and algorithms used by TSN switching devices to determine when, in what order, and how to send messages for different task flows. The core goal is to ensure metrics such as latency, bandwidth, real-time performance, and fairness.

[0079] Performance indicator data can refer to the quantitative evaluation data of the operation results of each scheduling strategy in the digital twin network model, which is used to analyze whether the network meets the industrial real-time communication requirements under the strategy.

[0080] First, the system obtains a unified system time base across the entire network through a completed time synchronization process, which serves as a unified time reference for all simulation behaviors. Subsequently, the system inputs various task parameters extracted from historical service data into the simulation model, including task periodicity, maximum delay tolerance, bandwidth requirements, jitter tolerance, and priority level. These parameters are used to define different types of task flow models and configure the generation behavior of virtual flows in the simulation. Next, load forecast results constructed based on historical load data and a unified timeline are input into the simulation system to generate a simulation of service traffic intensity within a specific time window within the future cycle, ensuring that the simulation accurately reflects the impact of dynamic load changes on the performance of the scheduling strategy.

[0081] The system then loads a preset list of scheduling policies, which includes several typical TSN scheduling methods (such as TAS, CBS, hybrid scheduling, and strict priority) and their scheduling parameter combinations. The simulation system calls the scheduling configuration module in the order of the policies in the list, loading the queue management logic, time slot configuration, gate control list (GCL), or credit calculation rules for each scheduling policy into the simulation scheduling engine.

[0082] During the simulation run phase, the TSN digital twin network model drives the event scheduler based on the input system time base. It simulates the forwarding of data packets from source nodes to destination nodes according to the network topology and clock synchronization configuration, and records the end-to-end transmission process of each task flow under various scheduling strategies. Throughout the entire process, the simulation engine monitors multiple performance indicators in real time and outputs performance data for each strategy after the simulation completes, including maximum end-to-end latency, jitter, packet loss rate, link utilization, time window misalignment, and priority default rate.

[0083] The training process of the preset TSN digital twin network model is:

[0084] A basic simulation framework is constructed based on real industrial network topologies, TSN protocol specifications, and device behavior standards. This framework should include virtual switching nodes, terminal nodes, link bandwidth configuration, clock synchronization logic (such as IEEE802.1AS), scheduler models (such as the Time-Aware Scheduler (TAS) and the Credit-Based Scheduler (CBS), and queue forwarding mechanisms. To ensure the model's generalization capabilities, the training dataset requires the collection of operational data from a variety of real-world industrial network configuration scenarios, including different types of task flows (periodic and bursty), scheduling configurations, and execution results under different bandwidth and latency constraints.

[0085] In the data generation stage, a large number of training samples are constructed using industrial simulation platforms (such as OMNeT++, NS-3 with NeSTiNg extension) or actual deployment platforms. Each sample consists of four parts: input features include system time reference (used to unify scheduling trigger time points), task parameters (such as cycle, delay tolerance, bandwidth requirements, etc.), scheduling strategy parameters (such as GCL cycle, time slot length, priority allocation rules) and load forecast results (bandwidth usage trend of task flows in the future); the corresponding labels are real or simulated performance indicator data generated under the policy configuration (such as end-to-end delay, maximum jitter, packet loss rate, resource occupancy, etc.).

[0086] The input and labels are then fed into the TSN digital twin model as training samples for modeling and training. The core of the model can be a hybrid model combining a graph neural network (GNN) and an attention mechanism: the GNN is used to model the network topology and multi-node scheduling coordination, while the attention mechanism is used to dynamically focus on local paths and task flows that significantly impact key performance indicators under specific scheduling strategies and traffic conditions. The model is trained using an end-to-end regression approach, with the goal of minimizing the error between the predicted performance indicator and the true indicator. A loss function such as the weighted mean squared error (WMSE) is used to emphasize the accuracy of predictions for key indicators such as latency.

[0087] The training process requires enhanced implementation using diverse policy configurations and multi-task workload scenarios, ensuring that the model not only fits known policy behaviors but also generalizes to less clearly defined policy combinations. In some scenarios, simulation-enhanced datasets or transfer learning can be used to further expand the model's adaptability.

[0088] After training, the TSN digital twin model can accept system time bases, task parameters, scheduling strategy lists, and load forecasts as input during actual deployment. Based on the learned relationship between scheduling behavior and network characteristics, it rapidly performs simulation predictions under various strategies and outputs performance metrics for different strategy combinations, providing data support and optimization for scheduling decisions. The model also supports incremental learning, periodically updating parameters based on field feedback data to ensure consistently high fit and simulation accuracy.

[0089] S106 , determining an optimal scheduling strategy based on the performance indicator data and the task parameters, and sending the optimal scheduling strategy to the TSN switching device to complete the scheduling control configuration of various task flows in the network.

[0090] The optimal scheduling strategy can be a scheduling scheme that can best meet the real-time, reliability and resource utilization efficiency requirements of all task flows under given network operating environment and task load conditions.

[0091] TSN switching devices can be Ethernet switching devices that support the TSN protocol family (IEEE802.1 series) and have deterministic forwarding capabilities. They are the key hardware infrastructure for achieving real-time and reliability assurance of TSN networks.

[0092] After completing simulations of multiple scheduling strategies, multi-dimensional performance indicator data corresponding to each preset scheduling strategy can be obtained, such as: end-to-end delay, maximum jitter, bandwidth occupancy, scheduling success rate (i.e., whether it can be transmitted on time), number of scheduling conflicts, GCL configuration complexity, resource utilization, etc. of each task flow. Subsequently, the system matches and verifies these performance indicators with the task parameters corresponding to each type of task flow (including: periodicity, maximum delay tolerance, bandwidth requirements, jitter tolerance, and priority level). For example, if the delay of a periodic task flow exceeds its maximum delay tolerance or the jitter exceeds the tolerance under a certain scheduling strategy, the strategy is determined to be unsatisfactory and is removed from the candidate set.

[0093] Next, the system performs a comprehensive scoring process on the remaining scheduling strategies. This can be done using multi-attribute decision-making models (such as TOPSIS, AHP, or weighted scoring). During this process, the system assigns weights to different performance indicators. For example, for high-real-time tasks, end-to-end latency and jitter are weighted higher; for high-throughput tasks, bandwidth utilization is weighted higher. The system calculates a standardized score based on the indicator values ​​for each scheduling strategy and performs a weighted aggregation to generate a comprehensive score list for all scheduling strategies. The highest-scoring scheduling strategy is selected as the optimal scheduling strategy.

[0094] Once the optimal scheduling strategy is determined, the system structures its scheduling content, mainly including: scheduling type (such as TAS, CBS), GCL scheduling table (defining the time slot switching sequence of each port), priority mapping table (such as the mapping relationship between PCP value and task in 802.1Q), port queuing rules and traffic shaping parameters (such as token bucket parameters of CBS), and packages the above configurations into configuration instructions through TSN control interface protocols (such as NETCONF, RESTCONF, OPC UA or dedicated TSN SDN control protocol).

[0095] These instructions are then sent by the system to each TSN switch. The device parses and loads the GCL table into the local scheduling control unit, configuring the local queue priority and gating scheduling mechanism. Under the synchronized clock reference, the TSN switch begins executing the optimal scheduling strategy, achieving precise timing and forwarding control for all task flows in the network, ensuring that all critical business flows are delivered on time within the specified time window, meeting the high reliability requirements of deterministic communication in industrial control networks.

[0096] In the embodiment of the present application, network node information of the industrial site is obtained, TSN topology deployment and configuration are performed according to the network node information, and an industrial network topology structure that supports the TSN protocol is obtained; the local system time information of each TSN network node in the industrial network topology structure is obtained, and the local system time information is accurately synchronized to obtain a unified system time base for the entire network; historical communication data and service characteristic data of various task flows in the industrial site are obtained, and feature extraction is performed on the historical communication data and service characteristic data to obtain task parameters of various task flows; wherein the task parameters include periodicity, maximum delay tolerance, bandwidth requirements, jitter tolerance, and priority, etc. Level; obtain historical load data of various task flows, time-align the historical load data according to the system time reference, construct a unified time series, input the unified time series into the preset prediction model, and obtain the load prediction results of various task flows; input the system time reference, task parameters, preset scheduling strategy list and load prediction results into the preset TSN digital twin network model, perform simulation of multiple scheduling strategies, and obtain performance indicator data under each scheduling strategy; determine the optimal scheduling strategy based on the performance indicator data and the task parameters, send the optimal scheduling strategy to the TSN switching device, and complete the scheduling control configuration of various task flows in the network. Through the above-mentioned TSN-based industrial real-time data transmission guarantee method, by integrating system time synchronization, task parameter modeling, load prediction and TSN digital twin simulation, intelligent evaluation and automatic optimization of scheduling strategies are realized, which not only significantly improves the real-time and deterministic communication capabilities of the industrial network, but also enhances resource utilization efficiency and the adaptability of scheduling decisions, ensuring that critical task flows are transmitted on time and reliably.

[0097] Based on the above technical solution, optionally, after completing the scheduling control configuration of various task flows in the network, the method further includes:

[0098] Obtaining actual operating status data of various task flows, and comparing and analyzing the actual operating status data with load forecast results;

[0099] If the deviation between the actual operating status data and the load forecast result exceeds the preset threshold, the preset feedback control mechanism will be executed to dynamically update the optimal scheduling strategy.

[0100] In this solution, the actual operation status data can be the operation data collected in real time from the TSN network during the actual operation of the task flow, which is used to reflect the actual load situation of the current task flow communication. It can include real-time bandwidth usage, actual message sending frequency and bytes, network transmission delay, link utilization, packet loss rate, and queue congestion information.

[0101] The preset threshold can be the judgment boundary set in advance by the system at the initial deployment to judge the accuracy of load prediction and the reliability of scheduling effect. It can include bandwidth prediction deviation threshold (such as ±10%), delay error threshold (such as ±5ms), jitter change rate threshold, task flow transmission success rate lower limit (such as 95%), and congestion count threshold.

[0102] The pre-set feedback control mechanism can be a closed-loop dynamic optimization strategy system designed in industrial TSN (time-sensitive networking) scheduling systems to address significant deviations between the actual operating status of task flows and predicted results. This ensures that the system can continuously adapt and optimize scheduling during operation. When a significant deviation between actual operating status data and load forecast results is detected, the system triggers a response based on a preset threshold. First, the prediction model is adaptively adjusted, such as through incremental learning, sliding window training, or fine-tuning model parameters to update the prediction results. Then, the scheduling strategy simulation is re-executed in the TSN digital twin network model, and the pre-set scheduling strategy list is re-evaluated based on the new predicted data to select a new optimal scheduling strategy. Finally, the updated optimal scheduling strategy is distributed to the TSN switching equipment in real time to complete the scheduling configuration switch. Time synchronization mechanisms or buffered scheduling windows are used to ensure a smooth transition of scheduling switches. The mechanism also includes strategies to suppress frequent switching, such as setting a minimum adjustment interval, a policy rollback mechanism, and policy version consistency management, to ensure stable and efficient system operation under dynamic changes.

[0103] The real-time collection mechanism deployed on TSN switching nodes and terminal devices can be used to obtain the actual operating status data of various task flows. This data includes key operating parameters such as the actual communication cycle (T_actual), average delay (D_avg_actual), maximum delay (D_max_actual), actual bandwidth occupancy (B_actual), packet loss rate (L_actual), and instantaneous jitter (J_actual) of each task flow. After being standardized and synchronized with the system time base, this data is compared item by item with the load forecast results output by the prediction model in the early stage. The load forecast results include parameters that correspond to actual values, such as the predicted cycle (T_pred), predicted bandwidth requirement (B_pred), predicted delay (D_pred), and predicted jitter range (J_pred). The comparison and analysis process uses defined preset thresholds to make deviation judgments. For example, the allowable error range is set as: |T_actual-T_pred|≤ε_T (such as 5% period error), |B_actual-B_pred|≤ε_B (such as ±10% bandwidth error), |D_max_actual-D_pred|≤ε_D (such as 3ms), |J_actual-J_pred|≤ε_J (such as 0.5ms). Only when any deviation exceeds these threshold limits will the system trigger the feedback control mechanism. The feedback control mechanism consists of two parts: First, the load prediction model is updated using an incremental learning strategy. In this case, the system introduces actual operating data as new samples and retrains or fine-tunes the prediction model parameters using a short sliding window (e.g., the last 5 minutes of data). Common algorithms include time series prediction networks such as LSTM and GRU. Second, the updated prediction values ​​are re-input into the TSN digital twin network model along with the current system time base and task parameters (such as priority P_i, maximum delay tolerance D_i, bandwidth requirement B_i, and periodicity T_i). A full simulation is then performed under a list of existing scheduling strategies (such as periodic scheduling, preemptive scheduling, gated scheduling, and bandwidth-aware scheduling). The performance metrics of each strategy under the new load conditions are evaluated, such as scheduling success rate S_i, average delay D_i_avg, packet loss rate L_i, and link utilization U_i. Finally, a weighted ranking is performed based on task parameters (especially priority and delay requirement) and performance metrics, and a new optimal scheduling strategy is selected and distributed to the TSN switching equipment through a centralized controller.

[0104] In this solution, by predicting parameters such as bandwidth, period, and delay of future task flows, possible congestion or conflicts can be identified in advance, thereby improving the system's forward-looking scheduling capabilities.

[0105] Based on the above technical solution, optionally, after completing the scheduling control configuration of various task flows in the network, the method further includes:

[0106] Determine the key task flow of each task flow according to the task parameters, obtain the data packet of the key task flow, perform frame replication and path distribution on the data packet of the key task flow, and obtain the data packet of multi-path redundant transmission;

[0107] Based on the data packets transmitted redundantly over multiple paths, data frame sequence number comparison and redundant data elimination are performed to obtain fault-tolerant data frames, which are then sent to a control system for task instruction execution.

[0108] In this solution, critical task flows refer to those with extremely high requirements for real-time performance, reliability, and security within industrial TSN networks. These tasks typically carry core production control instructions, such as PLC control instructions, industrial robot motion instructions, and safety control commands. Their characteristic parameters (such as maximum delay tolerance, jitter tolerance, and priority level) are highly sensitive to task parameters. Transmission errors or delays can lead to system failures or safety risks.

[0109] Mission-critical data packets are frame- or packet-level data units generated by mission-critical flows during communications, carrying specific service data and control information. These packets have specific periodicity, bandwidth requirements, and priority identifiers, and must be reliably delivered to the target device within a specified time window.

[0110] Multipath redundant transmission refers to a collection of copies of critical mission-critical data packets that are transmitted in parallel across multiple different links in the TSN network after frame duplication. This is a key method for implementing the "frame duplication and elimination" feature of TSN's fault-tolerance mechanism, ensuring that data transmitted along the secondary path can still be delivered on time even if the primary path fails or is disrupted.

[0111] A fault-tolerant data frame refers to the final valid frame generated by the receiver (such as a TSN switch or controller) after serial number identification, duplicate frame removal, and sequential reassembly of redundant data packets across multiple paths. Fault-tolerant data frames are reliable data frames that have been processed through redundancy mechanisms to eliminate the effects of faults and can be directly used for the execution and feedback of industrial tasks.

[0112] A control system refers to the system units connected to the TSN network that are responsible for actual industrial equipment control and task scheduling, such as PLCs (programmable logic controllers), DCSs (distributed control systems), motion controllers, and industrial servers. Fault-tolerant data frames are ultimately delivered to the control system, driving the equipment to perform precise control operations.

[0113] To ensure the real-time and reliability of critical control instructions in industrial TSN networks, all task flows must first be classified and prioritized based on task parameters (including periodicity, maximum delay tolerance, bandwidth requirements, jitter tolerance, and priority level). By constructing a priority discrimination model (such as the AHP method based on weighted multi-criteria decision-making or the threshold rule method), task flows with low delay tolerance, high jitter requirements, short data cycles, and high priority levels can be identified. These are defined as critical task flows. Next, the data packets corresponding to these critical task flows are captured in real time within the industrial communication system and processed by the Frame Replication and Elimination for Reliability (FRER) mechanism. Specifically, critical data packets undergo frame replication at the source node or ingress TSN switch. Based on predefined path distribution strategies (such as those based on minimum path conflicts, maximum link bandwidth redundancy, or historical reliability indicators), each replicated frame is dynamically allocated two or more independent transmission paths. These replicated frames are then sent in parallel along different paths within the TSN network, generating data packets with multi-path redundant transmission. Frame aggregation and comparison operations are performed based on the data frame sequence number embedded in each frame of data packets transmitted redundantly over multiple paths. Each arriving data copy is placed in a buffer indexed by the sequence number. The system monitors in real time whether data frames on multiple paths have the same sequence number. Once multiple copies with the same sequence number are identified, the system initiates redundancy elimination logic: using a "first arrival first" or "verification result first" strategy, the copy with correct content and minimal latency is selected as the valid frame, and the remaining copies are immediately discarded. The system can also optionally perform content consistency checks on these copies to further enhance error detection capabilities. The resulting single valid frame is marked as a fault-tolerant data frame.

[0114] These fault-tolerant data frames are then sent to industrial control systems (such as PLC, DCS, motion controllers, etc.) to execute related task instruction control logic, thereby ensuring the accurate, safe and high-real-time execution of key control instructions.

[0115] In this solution, by embedding data frame sequence numbers in data packets transmitted redundantly over multiple paths, the system can accurately identify and eliminate duplicate frames at the receiving end, retaining only the first valid copy, thereby improving the reliability and fault tolerance of mission-critical flows in industrial networks.

[0116] Based on the above technical solution, optionally, after completing the scheduling control configuration of various task flows in the network, the method further includes:

[0117] Obtaining collaborative control instructions, and extracting periodic scheduling requirements, time sensitivity levels, and resource demand information of various task flows based on the collaborative control instructions;

[0118] Mapping the periodic scheduling requirements to corresponding time windows, mapping the time sensitivity levels to corresponding time slot resource configurations, and mapping the resource demand information to network transmission path configurations;

[0119] A synchronous scheduling strategy is generated according to the time window, time slot resource configuration and network transmission path configuration, and various task flows are scheduled and executed according to the synchronous scheduling strategy.

[0120] In this solution, collaborative control instructions refer to control commands issued by a higher-level control system (such as an industrial control master station or a dispatch control center) to coordinate the synchronous execution of multiple task flows. These instructions typically include task dependencies, scheduling priorities, and synchronization requirements, ensuring that multiple tasks or devices operate in coordination according to established time and resource policies.

[0121] Periodic scheduling requirements can refer to the frequency and periodicity of repetitive execution of various task flows over time, such as a task being triggered every 5ms or 10ms. This reflects the periodic scheduling characteristics of tasks and serves as the basis for setting time windows.

[0122] The time sensitivity level (TSL) is a metric that measures a task's tolerance for time delay, transmission jitter, and scheduling accuracy. It is typically categorized as high, medium, or low. For example, motion control tasks are classified as highly time-sensitive, while general data collection tasks might be classified as low. This metric determines scheduling urgency and priority.

[0123] Resource requirement information can include parameters such as the network bandwidth, routing path, buffer capacity, and computing resources required for the task flow during transmission. For example, a high-definition video streaming task may require a sustained bandwidth of ≥10 Mbps, while a control instruction transmission may only require a few hundred bytes.

[0124] A time window refers to a transmission time period divided within a network-wide unified time base to meet periodic scheduling requirements. Each periodic task is assigned to one or more dedicated time windows to ensure timely triggering and avoid conflicts. For example, the time window for a 5ms periodic task might be a 1ms segment every 5ms.

[0125] Time slot resource allocation is the result of fine-grained division and allocation of time resources within a scheduling cycle based on time sensitivity levels. Highly sensitive tasks are assigned earlier time slots with less jitter. For example, within a scheduling cycle, the first few microseconds might be dedicated to time-sensitive tasks.

[0126] Network transmission path configuration can select and preset the optimal transmission path for the task flow based on resource demand information and the network topology of TSN (time-sensitive network), which may include primary and backup paths, bandwidth reservation, path conflict avoidance, etc.

[0127] A synchronous scheduling strategy can be a unified scheduling execution plan for all task flows, combining time windows, time slot configurations, and transmission paths. This strategy, in the form of a schedule or scheduling rules, controls when and where tasks are transmitted, along specific times and paths, with what bandwidth, and whether redundancy mechanisms are enabled, thereby ensuring efficient and orderly execution of tasks according to collaborative requirements.

[0128] Collaborative control instructions, which contain the scheduling and communication requirements of multiple tasks, can be obtained through the master scheduling system, edge devices, or control servers. The system uses protocol parsing and structured semantic analysis technology to extract the periodic scheduling requirements, time sensitivity level, and resource requirement information of each type of task flow from the collaborative control instructions. Specifically, the periodic scheduling requirement refers to the triggering frequency of the task flow (such as execution every 5ms), which is used to describe the periodic scheduling characteristics of the task; the time sensitivity level is used to measure the tolerance of the task to network transmission delay, usually divided by priority level (such as TSN priority 0 to 7); resource requirement information refers to the task's demand for network resources, including the required bandwidth (such as 50Mbps), the required transmission path (such as through switching nodes A, B, C), and the transmission reliability level. After extracting the above parameters, the system first maps the periodic scheduling requirements to the corresponding time window through a unified network clock mechanism. For example, a task with a period of 10ms is assigned to the 0th to 1ms period of each scheduling cycle for execution. Subsequently, a time sensitivity-priority mapping table is used to convert time sensitivity levels into specific time slot resource configurations. Tasks of different levels are scheduled to different time slot resources, and scheduling protection mechanisms (such as GuardBand isolation zones) are configured for time-sensitive tasks. Simultaneously, network topology awareness and constrained path search techniques are employed, in conjunction with resource demand information, to determine the network transmission path configuration that meets task requirements. Control strategies such as path identification, traffic shaping, and bandwidth reservation are then configured. Based on these time windows, time slot resource configurations, and network transmission path configurations, the system automatically generates a synchronous scheduling policy for network-wide task synchronization. This policy includes key scheduling parameters such as task scheduling start time, duration, allocation channel, data frame transmission time slot, and transmission path. Ultimately, this synchronous scheduling policy is distributed to the relevant switching devices and terminal nodes in the TSN network, ensuring time synchronization, high priority assurance, and resource conflict avoidance for various task flows in a multi-source, heterogeneous industrial network.

[0129] In this solution, by parsing collaborative control instructions into specific time windows, time slot resources and network path configurations, a synchronous scheduling strategy is generated, which realizes the precise scheduling and resource collaboration of task flows in industrial networks.

[0130] Based on the above technical solution, optionally, after completing the scheduling control configuration of various task flows in the network, the method further includes:

[0131] Obtaining the operating status log data during the operation of the industrial network and the data statistical information reported by the network devices, performing multi-dimensional fusion processing on the operating status log data and the data statistical information to obtain a fused network status assessment data set;

[0132] The network status evaluation data set is processed using a statistical analysis method and a rule-based discrimination algorithm to obtain network operation performance index results.

[0133] In this context, an industrial network refers to the communication network used in industrial automation systems, connecting sensors, actuators, controllers, PLCs, industrial hosts, and other devices to exchange data and transmit control information. It supports communication protocols that require high real-time performance, determinism, and reliability.

[0134] Operational status log data refers to the behavioral data automatically recorded by industrial network devices (such as switches, routers, and terminal nodes) during operation. It usually includes timestamp information such as task flow transmission records, packet loss records, device exception logs, interface online and offline times, forwarding delays, and queuing delays.

[0135] Network devices can refer to hardware involved in data forwarding, routing, management, or terminal communication in industrial networks, including TSN switches, industrial gateways, and terminal nodes (such as PLCs or embedded devices). These devices report performance statistics periodically or in an event-driven manner.

[0136] Data statistics can be operational indicator data collected and reported by network devices in real time or periodically, including throughput, packet forwarding rate, link utilization, delay, packet loss rate, clock synchronization error, resource occupancy, etc., reflecting the current network performance status.

[0137] The network status assessment dataset can be a structured data set formed by integrating and organizing the operation status log data and data statistical information in the time dimension, device dimension, task flow dimension, etc., which is used to reflect the comprehensive operation status of the entire industrial network at present or in a certain period of time.

[0138] Statistical analysis methods can include time series analysis, mean-variance calculation, correlation analysis, sliding window averaging, anomaly detection, etc., which are used to quantify data trends and fluctuations.

[0139] The rule-based discrimination algorithm can be based on predefined business rules or expert system rule bases (such as "delays exceeding Xms are considered abnormal") to perform conditional judgment and classification on the data set to assist in identifying faults or potential risks.

[0140] The network operation performance indicator results can be the result output after statistical analysis and rule judgment processing, usually including network delay indicators, packet loss rate, clock synchronization accuracy, bandwidth utilization, task flow transmission stability level, equipment health score, etc., which can be used for performance evaluation, operation and maintenance decision-making or dynamic scheduling optimization.

[0141] A distributed collection system can capture real-time operational status log data and statistical data reported by network devices during industrial network operation. Operational status log data can be continuously collected by log agents (e.g., Fluentd or Logstash-based collection modules) deployed at key network nodes (e.g., switches, controllers, and terminal devices). This data includes data frame transmission timestamps, path hop counts, task flow execution results, retransmission records, and exception event codes. Statistical information is periodically reported via the industrial network devices' SNMP interfaces, NETCONF interfaces, or proprietary APIs. This information includes real-time statistical metrics such as port bandwidth utilization, queue length, forwarding delay, synchronization error, link load, retransmission count, and packet loss rate. Processing modules at edge computing nodes or central servers then perform multi-dimensional fusion processing on these two types of data. This includes timestamp alignment based on a unified time base (e.g., PTP clock), correlation matching of task flow identifiers, and the construction of fusion matrices across node, link, and service type dimensions. Data cleaning, normalization, and feature mapping are then employed to generate a unified, fused network status assessment dataset. This dataset, centered around task flows, correlates the performance data of each node on these flows, forming a complete network status map. This fused dataset is intelligently processed using a combination of statistical analysis methods and rule-based identification algorithms. Statistical analysis uses techniques such as mean / variance calculation within a sliding time window, trendline fitting, cluster anomaly detection (such as DBSCAN or LOF), and periodic fluctuation modeling to analyze the stability and volatility of key indicators. Rule-based identification uses pre-set network operation rule templates (e.g., "Packet loss rate >1% is an anomaly" and "Sync error exceeding ±100ns requires a warning") to classify and rank each task flow and node status. Output network performance indicators include task flow stability ratings, node health scores, link bottleneck identification results, service latency assessments, and lists of abnormal alarm events, providing decision-making support for dynamic network scheduling, fault-tolerant switching, and fault early warning. The entire process relies on a time series database (such as InfluxDB), a metric calculation engine (such as Prometheus), and data visualization tools (such as Grafana) for real-time visualization and response.

[0142] In this solution, by obtaining operation status log data and data statistical information and performing integrated analysis, the operation status of the industrial network can be comprehensively, real-time, and accurately grasped, and potential faults and performance bottlenecks can be discovered in a timely manner.

[0143] Figure 2 This is a flow chart of a method for ensuring industrial real-time data transmission based on TSN provided in an embodiment of the present disclosure. The method may include the following steps:

[0144] S201, obtaining network node information of an industrial site, performing TSN topology deployment and configuration according to the network node information, and obtaining an industrial network topology structure supporting the TSN protocol.

[0145] S202, obtaining local system time information of each TSN network node in the industrial network topology, accurately synchronizing the local system time information, and obtaining a unified system time reference for the entire network.

[0146] S203, obtaining historical communication data and service characteristic data of various task flows in the industrial site, performing feature extraction on the historical communication data and service characteristic data, and obtaining task parameters of the various task flows; wherein the task parameters include periodicity, maximum delay tolerance, bandwidth requirement, jitter tolerance, and priority level.

[0147] S204, obtaining historical load data of various task flows, time-aligning the historical load data according to the system time reference, constructing a unified time series, inputting the unified time series into a preset prediction model, and obtaining load prediction results of various task flows.

[0148] S205: Input the system time base, task parameters, preset scheduling strategy list and load forecast results into the preset TSN digital twin network model, perform simulation of multiple scheduling strategies, and obtain performance indicator data under each scheduling strategy.

[0149] S206 , determining an optimal scheduling strategy based on the performance indicator data and the task parameters, and sending the optimal scheduling strategy to the TSN switching device to complete the scheduling control configuration of various task flows in the network.

[0150] S207, obtain the load prediction results, scheduling resource allocation parameters and network bandwidth configuration data of each type of task flow within the preset evaluation time window, and calculate the predicted sending load of each type of task flow in each scheduling cycle within the preset evaluation time window based on the load prediction results and scheduling resource allocation parameters.

[0151] The preset evaluation window is a time period defined by the system for load forecasting and bandwidth allocation. For example, it can be set to 5 minutes or 30 scheduling cycles in the future. The system will evaluate the load changes, resource allocation, and available bandwidth of the task flow within this time period, enabling more accurate scheduling decisions.

[0152] Scheduling resource allocation parameters refer to the resource scheduling ratio parameters pre-set by the system for various task flows, typically including the number of allocated time slots, priority level, maximum allowable send load, and send period. These parameters are output by the scheduling policy generator or TSN digital twin model and are used to guide the transmission behavior of task flows in the network.

[0153] Network bandwidth configuration data can be the bandwidth resource allocation status of each link or communication path in the current industrial network, such as the total bandwidth of each link, reserved bandwidth, remaining available bandwidth, and the bandwidth ratio reserved for different priority flows. It reflects the resource carrying capacity of the network within the scheduling cycle.

[0154] The scheduling period is the time unit used by the scheduling system to schedule task flows. For example, if scheduling is performed every 1ms, the scheduling period is 1ms. The entire evaluation time window is divided into several scheduling periods to facilitate fine-grained resource prediction and allocation.

[0155] The predicted send load refers to the amount of data (usually expressed in bytes / bits) that a task flow is expected to send during each scheduling cycle, calculated based on historical load data and prediction models (such as LSTM and time series regression). This is an important indicator for determining whether there is overload risk.

[0156] Load forecasts, scheduling resource allocation parameters, and current network bandwidth configuration data for various task flows within a preset evaluation time window can be obtained from the scheduling control center or upper-level scheduling system. Based on each task flow's historical load data and current business trends, the system uses a time series prediction model (such as an LSTM neural network or sliding window weighted average method) to output the predicted sending load for each scheduling cycle within the evaluation time window. Specifically, the system collects historical load data for each task flow from previous scheduling cycles and combines it with current business growth trends or fluctuations to form a continuous load change sequence. This data is processed using a time series prediction model. If a long-short-term memory neural network model is used, the system uses historical load data from multiple consecutive cycles as input, trains the model to identify temporal dependencies and trend characteristics, and predicts load changes for subsequent scheduling cycles. If a sliding window weighted average method is used, a fixed number of historical load values ​​are assigned different weights to each cycle, and the weighted average is calculated to produce the predicted sending load for each subsequent scheduling cycle. Ultimately, the system outputs the predicted load for each task flow type across each scheduling cycle within the entire evaluation time window for reference in scheduling and resource allocation. Subsequently, based on the scheduling resource allocation parameters assigned to each task flow type, such as cycle length, priority level, and bandwidth share, the predicted load is mapped to the available scheduling resources. The target sending volume for the task flow within each scheduling cycle is calculated, ensuring that it matches the predicted value with the bandwidth and cycle resources, thus achieving refined resource management driven by prediction.

[0157] S208 , obtaining scheduling parameters of each scheduling period, and calculating the allocatable bandwidth load upper limit of each scheduling period of each task flow within a preset evaluation time window according to the network bandwidth configuration data and the scheduling parameters.

[0158] Scheduling parameters refer to the scheduling control parameters set by the system for the task flow in each scheduling cycle, including available sending time slots, maximum transmittable data volume, task flow scheduling priority, path selection strategy, etc. They guide the specific scheduling execution actions of each cycle.

[0159] The allocatable bandwidth load cap refers to the maximum permissible bandwidth load limit calculated by the system for each task flow during each scheduling cycle, taking into account factors such as the current link bandwidth, task flow priority, and allocated resources. This limit prevents link overload and ensures reliable transmission of high-priority flows.

[0160] The execution plans for multiple scheduling cycles can be preset through the scheduling control strategy configuration mechanism in the control architecture of the industrial network. In actual operation, the scheduling configuration information in the network is collected in real time through the network control node (such as a centralized master controller or an edge scheduling node) to obtain the key scheduling parameters of each scheduling cycle. These scheduling parameters specifically include the scheduling priority of various task flows (which can be determined by task level division rules or based on task delay sensitivity calculations), the transmission time slice or time slot allocated to each task flow (based on the real-time nature and bandwidth requirements of the task, such as time slot rotation, earliest deadline priority, etc.), and the start and end time allowed for transmission of each task flow (based on the task release time and the latest completion time). After obtaining the scheduling parameters, the system enters the next stage, which is to use network topology information and real-time link information to obtain the network bandwidth configuration data of the current network. This step is accomplished through a network status awareness mechanism, utilizing the following technologies: a real-time link bandwidth monitoring algorithm, including end-to-end link capacity measurement based on active probing (e.g., sending specific probe packets and analyzing their return times); link load statistical analysis methods, such as using a sliding window to calculate average and peak traffic flow within the current timeframe to estimate available bandwidth; and a link congestion assessment model, which combines historical bandwidth utilization and latency trends to determine link congestion. The system then divides the entire pre-set evaluation window into multiple consecutive scheduling cycles. For each cycle, combining the scheduling parameters obtained in the previous step with the network bandwidth configuration data, it calculates the allocatable bandwidth load cap for each task flow within that cycle. This calculation step uses the following methods: path analysis and mapping: determining the transmission path of each type of task flow (i.e., the sequence of network node links that the task passes through from the source node to the target node); link bandwidth cross-constraint analysis: traversing each link on the path and taking the minimum available bandwidth on the path as the bandwidth bottleneck; task bandwidth demand assessment: combining the task data size and transmission cycle length in the scheduling parameters to calculate the theoretical bandwidth requirement of the task; bandwidth load matching and limit correction: comparing the theoretical bandwidth requirement with the path bottleneck bandwidth. If the bottleneck bandwidth is insufficient, a weighted allocation algorithm (such as proportional allocation based on task priority weighting) is used to calculate the adjusted allocable bandwidth load upper limit. Ultimately, the system outputs the allocable bandwidth load upper limit for each task flow in each scheduling cycle.

[0161] In this embodiment, the load demand and network bandwidth resource matching of the task flow in each future scheduling period can be evaluated in advance, and the resource allocation strategy can be dynamically adjusted to avoid scheduling conflicts and task delays caused by load bursts or insufficient bandwidth, thereby improving network transmission efficiency and system operation stability.

[0162] Based on the above technical solution, optionally, after calculating the upper limit of the allocatable bandwidth load of each task flow in each scheduling period within the preset evaluation time window, the method further includes:

[0163] If there is a task flow whose predicted sending load exceeds the upper limit of the allocable bandwidth load, the corresponding task flow will be used as the target task flow;

[0164] Generate delay warning information based on the predicted sending load of the target task flow, the upper limit of the allocatable bandwidth load, and the scheduling resource allocation parameters.

[0165] In this solution, target task flows are those whose predicted transmit load exceeds the upper limit of the allocable bandwidth within a specific scheduling cycle. Due to insufficient bandwidth resources, these task flows may face risks such as increased transmission latency or data congestion, and are therefore identified as requiring focused monitoring and action.

[0166] Latency warning information is generated based on an analysis of the target task flow's predicted send load, the upper limit of allocable bandwidth, and scheduling resource allocation parameters. This information alerts the system to potential risks of insufficient scheduling resources or network congestion, helping to proactively adjust resources or optimize scheduling strategies to reduce the probability of task failures due to latency.

[0167] When the system detects that a task flow's predicted send load exceeds the corresponding allocatable bandwidth load limit, it identifies it as a target task flow and conducts a detailed analysis of its potential delay risk. The system extracts relevant parameters for the target task flow: the predicted send load for the current scheduling cycle, the corresponding allocatable bandwidth load limit, the task's scheduling resource allocation parameters (such as task priority, task periodicity, and latency tolerance threshold), and the task's own time sensitivity level. Next, the system performs multi-dimensional calculations based on a predefined evaluation model. The first step is to calculate the task's bandwidth shortage ratio: the difference between the predicted send load and the allocatable bandwidth load limit divided by the limit, quantifying the current bandwidth resource shortage. The second step is to assess the task's congestion impact index, which is weighted by combining historical traffic trends with the competition for shared resources among other task flows in the current cycle. The third step, based on the time sensitivity level, determines the task's tolerance for scheduling delays. For example, high-level control tasks often require millisecond-level response times, and scheduling delays can directly impact system stability. The system inputs the evaluation indicators of each of these dimensions into a rule-based judgment engine (or a rule-based scoring model), quantifying the latency risk level into three categories, such as "low risk," "medium risk," and "high risk." It also generates latency warnings for high-risk task flows. This warning information includes, but is not limited to, the task flow identifier, the calculated bandwidth gap percentage, the latency risk level, the corresponding scheduling cycle number, and the predicted worst-case transmission delay range. This information enables subsequent resource scheduling systems or operations personnel to dynamically respond with measures such as scheduling optimization, task rescheduling, or resource expansion, effectively preventing critical task delays caused by network congestion.

[0168] The training process of the set evaluation model is as follows: the set evaluation model is mainly trained through supervised learning, and its training data comes from the operation records of a large number of historical task flows, including the sending load, available bandwidth, task priority, scheduling resource allocation, time sensitivity level, actual transmission delay results, etc. of each task flow in the historical scheduling cycle. During the training process, these multi-dimensional feature data are first cleaned, normalized and feature-engineered to extract key input variables; then, whether the historical task flow has serious delays (such as exceeding the set delay threshold) is used as the target label to construct a classification model or regression model for the delay risk prediction problem. Commonly used models include decision trees, random forests, gradient boosting trees, support vector machines or neural networks. Through cross-validation and tuning of hyperparameters, an evaluation model is finally obtained that can predict the delay risk level of the task flow based on real-time input. After deployment, the model can run in real time in each scheduling cycle to predict the delay risk level of the detected target task flow.

[0169] The rule-based judgment engine is constructed and trained using a rule extraction and weight learning method based on an expert knowledge base and historical operating patterns. Domain experts first define a series of judgment rules based on their network scheduling experience and business characteristics, such as "If the bandwidth gap ratio exceeds 30% and the task priority is high, mark it as high risk." These rules are then input into the system as the initial rule set. The system then performs offline simulation playback on a large amount of historical data. By comparing the rule outputs with actual latency, the triggering conditions and weights of each rule are adjusted to improve accuracy and generalization. Furthermore, fuzzy logic is combined with a rule scoring mechanism to comprehensively score task flows that trigger multiple rules simultaneously, further quantifying the latency risk level. Ultimately, the rule-based judgment engine can quickly determine the risk level based on the current task parameters and output structured latency warning information to support the dynamic decision-making of the scheduling system.

[0170] In this solution, by combining historical operation data with real-time scheduling parameters, an evaluation model or rule judgment engine is used to identify and warn of delay risks in task flows in advance. This can locate potential bottlenecks before problems occur, adjust resource scheduling strategies in a timely manner, and avoid serious transmission delays caused by insufficient bandwidth.

[0171] Based on the above technical solution, optionally, after calculating the upper limit of the allocatable bandwidth load of each task flow in each scheduling period within the preset evaluation time window, the method further includes:

[0172] If there is no task flow whose predicted sending load exceeds the upper limit of the allocable bandwidth load, the preset evaluation time window is updated every time the preset evaluation interval is reached, and the load prediction results, scheduling resource allocation parameters and network bandwidth configuration data of each type of task flow within the preset evaluation time window are re-obtained. Based on the load prediction results and scheduling resource allocation parameters, the predicted sending load of each type of task flow in each scheduling period within the preset evaluation time window is recalculated;

[0173] The scheduling parameters of each scheduling period are re-acquired, and the allocatable bandwidth load upper limit of each scheduling period of each task flow within the preset evaluation time window is recalculated according to the network bandwidth configuration data and the scheduling parameters.

[0174] In this solution, the preset evaluation interval can be a fixed time period set by the system to implement periodic prediction and evaluation, which is used to trigger a new round of task flow prediction and network resource evaluation. This evaluation interval can be configured based on the dynamic nature of the network and the real-time requirements of the task. For example, it can be set to 5 seconds, 30 seconds, or 1 minute to ensure real-time system response while taking into account the consumption of computing resources.

[0175] The system is equipped with a high-precision timer or task scheduling mechanism. When the timer reaches a complete evaluation interval, the system enters a new prediction-evaluation process. First, the system updates the preset evaluation time window based on the time length set by sliding forward at the current moment. For example, the start and end time of the original evaluation window are shifted forward by the length of an evaluation interval, thereby forming a new time window coverage range. Then, the system re-circulates to obtain the load prediction results, scheduling resource allocation parameters and network bandwidth configuration data of various task flows within the preset evaluation time window. Based on the load prediction results and scheduling resource allocation parameters, the predicted sending load of various task flows in each scheduling cycle within the preset evaluation time window is recalculated; the scheduling parameters of each scheduling cycle are re-obtained, and based on the network bandwidth configuration data and scheduling parameters, the allocable bandwidth load upper limit of various task flows in each scheduling cycle within the preset evaluation time window is recalculated.

[0176] In this solution, by continuously updating the evaluation time window within each preset evaluation interval and dynamically recalculating the predicted sending load and allocable bandwidth limit of the task flow, the system can achieve real-time monitoring of the network status and forward-looking scheduling optimization, thereby discovering potential load change trends in advance.

[0177] Figure 3 This is a schematic block diagram of a TSN-based industrial real-time data transmission assurance system provided in an embodiment of the present disclosure. The system includes:

[0178] The TSN network topology deployment module 301 is used to obtain network node information of the industrial site, perform TSN topology deployment and configuration according to the network node information, and obtain an industrial network topology structure that supports the TSN protocol;

[0179] The system time synchronization module 302 is used to obtain the local system time information of each TSN network node in the industrial network topology, accurately synchronize the local system time information, and obtain a unified system time reference for the entire network;

[0180] The task parameter extraction module 303 is used to obtain historical communication data and service characteristic data of various task flows in the industrial site, perform feature extraction on the historical communication data and service characteristic data, and obtain task parameters of the various task flows; wherein the task parameters include periodicity, maximum delay tolerance, bandwidth requirement, jitter tolerance, and priority level;

[0181] The load prediction module 304 is used to obtain historical load data of various task flows, time-align the historical load data according to the system time reference, construct a unified time series, and input the unified time series into a preset prediction model to obtain load prediction results for various task flows;

[0182] The TSN simulation module 305 is used to input the system time base, task parameters, preset scheduling strategy list and load forecast results into the preset TSN digital twin network model, perform simulations of multiple scheduling strategies, and obtain performance indicator data under each scheduling strategy;

[0183] The scheduling optimization module 306 is used to determine the optimal scheduling strategy based on the performance indicator data and the task parameters, and send the optimal scheduling strategy to the TSN switching device to complete the scheduling control configuration of various task flows in the network.

[0184] Figure 4 A schematic block diagram of an electronic device 400 that can be used to implement an embodiment of the present disclosure is shown, including a processor 401, a memory 402, and a program or instruction stored in the memory 402 and executable on the processor 401. When the program or instruction is executed by the processor 401, each process of the above-mentioned TSN-based industrial real-time data transmission assurance method embodiment is implemented, and the same technical effect can be achieved. To avoid repetition, it is not described again here.

[0185] It should be noted that the electronic devices in the embodiments of the present application include the mobile electronic devices and non-mobile electronic devices mentioned above.

[0186] It should be noted that, in this article, the terms "comprise", "include" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or system comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or system. In the absence of further restrictions, an element defined by the statement "comprises a ..." does not exclude the presence of other identical elements in the process, method, article or system comprising the element. In addition, it should be noted that the scope of the methods and systems in the embodiments of the present application is not limited to performing functions in the order shown or discussed, and may also include performing functions in a substantially simultaneous manner or in the opposite order according to the functions involved. For example, the described method may be performed in an order different from that described, and various steps may also be added, omitted, or combined. In addition, the features described with reference to certain examples may be combined in other examples.

[0187] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus the necessary general hardware platform, and of course can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art can be embodied in the form of a computer software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), including a number of instructions for enabling a terminal (which can be a mobile phone, computer, server, or network device, etc.) to execute the methods described in each embodiment of the present application.

[0188] The embodiments of the present application are described above in conjunction with the accompanying drawings, but the present application is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of this application, ordinary technicians in this field can also make many forms without departing from the purpose of this application and the scope of protection of the claims, all of which are within the protection of this application.

[0189] The above are only preferred embodiments of the present application and the technical principles employed. The present application is not limited to the specific embodiments described herein, and various obvious changes, readjustments, and substitutions that are possible for those skilled in the art will not depart from the scope of protection of the present application. Therefore, although the present application has been described in more detail through the above embodiments, the present application is not limited to the above embodiments and may include more other equivalent embodiments without departing from the concept of the present application. The scope of the present application is determined by the scope of the claims.

Claims

1. A TSN-based industrial real-time data transmission guarantee method, characterized in that: The method comprises: Obtain network node information at the industrial site, perform TSN topology deployment and configuration based on the network node information, and obtain an industrial network topology structure that supports the TSN protocol; Obtain the local system time information of each TSN network node in the industrial network topology, accurately synchronize the local system time information, and obtain a unified system time benchmark for the entire network; Acquiring historical communication data and service characteristic data of various task flows in the industrial site, performing feature extraction on the historical communication data and service characteristic data, and obtaining task parameters of the various task flows; wherein the task parameters include periodicity, maximum delay tolerance, bandwidth requirement, jitter tolerance, and priority level; Obtain historical load data of various task flows, time-align the historical load data according to the system time base, construct a unified time series, input the unified time series into a preset prediction model, and obtain load prediction results for various task flows; Input the system time base, task parameters, preset scheduling strategy list, and load forecast results into the preset TSN digital twin network model, perform simulations of multiple scheduling strategies, and obtain performance indicator data under each scheduling strategy; An optimal scheduling strategy is determined based on the performance indicator data and the task parameters, and the optimal scheduling strategy is sent to the TSN switching device to complete the scheduling control configuration of various task flows in the network.

2. The method according to claim 1, characterized in that in, After completing the scheduling control configuration of various task flows in the network, the method further includes: Obtaining actual operating status data of various task flows, and comparing and analyzing the actual operating status data with load forecast results; If the deviation between the actual operating status data and the load forecast result exceeds the preset threshold, the preset feedback control mechanism will be executed to dynamically update the optimal scheduling strategy.

3. The method according to claim 1, characterized in that in, After completing the scheduling control configuration of various task flows in the network, the method further includes: Determine the key task flow of each task flow according to the task parameters, obtain the data packet of the key task flow, perform frame replication and path distribution on the data packet of the key task flow, and obtain the data packet of multi-path redundant transmission; Based on the data packets transmitted redundantly over multiple paths, data frame sequence number comparison and redundant data elimination are performed to obtain fault-tolerant data frames, which are then sent to a control system for task instruction execution.

4. The method according to claim 1, wherein in, After completing the scheduling control configuration of various task flows in the network, the method further includes: Obtaining collaborative control instructions, and extracting periodic scheduling requirements, time sensitivity levels, and resource demand information of various task flows based on the collaborative control instructions; Mapping the periodic scheduling requirements to corresponding time windows, mapping the time sensitivity levels to corresponding time slot resource configurations, and mapping the resource demand information to network transmission path configurations; A synchronous scheduling strategy is generated according to the time window, time slot resource configuration and network transmission path configuration, and various task flows are scheduled and executed according to the synchronous scheduling strategy.

5. The method according to claim 1, wherein in, After completing the scheduling control configuration of various task flows in the network, the method further includes: Obtaining the operating status log data and data statistical information reported by network devices during the operation of the industrial network, performing multi-dimensional fusion processing on the operating status log data and data statistical information to obtain a fused network status assessment data set; The network status evaluation data set is processed using a statistical analysis method and a rule-based discrimination algorithm to obtain network operation performance index results.

6. The method according to claim 1, characterized in that in, After completing the scheduling control configuration of various task flows in the network, the method further includes: Obtain load prediction results, scheduling resource allocation parameters, and network bandwidth configuration data for each type of task flow within a preset evaluation time window, and calculate the predicted sending load of each type of task flow in each scheduling cycle within the preset evaluation time window based on the load prediction results and scheduling resource allocation parameters; Obtain scheduling parameters for each scheduling period, and calculate the allocatable bandwidth load upper limit of each scheduling period for each type of task flow within a preset evaluation time window based on the network bandwidth configuration data and the scheduling parameters.

7. The method according to claim 6, characterized in that in, After calculating the allocatable bandwidth load upper limit of each scheduling period for each type of task flow within a preset evaluation time window, the method further includes: If there is a task flow whose predicted sending load exceeds the upper limit of the allocable bandwidth load, the corresponding task flow will be used as the target task flow; Generate delay warning information based on the predicted sending load of the target task flow, the upper limit of the allocatable bandwidth load, and the scheduling resource allocation parameters.

8. The method according to claim 6, characterized in that in, After calculating the allocatable bandwidth load upper limit of each scheduling period for each type of task flow within a preset evaluation time window, the method further includes: If there is no task flow whose predicted sending load exceeds the upper limit of the allocable bandwidth load, the preset evaluation time window is updated every time the preset evaluation interval is reached, and the load prediction results, scheduling resource allocation parameters and network bandwidth configuration data of each type of task flow within the preset evaluation time window are re-obtained. Based on the load prediction results and scheduling resource allocation parameters, the predicted sending load of each type of task flow in each scheduling period within the preset evaluation time window is recalculated; The scheduling parameters of each scheduling period are re-acquired, and the allocatable bandwidth load upper limit of each scheduling period of each task flow within the preset evaluation time window is recalculated according to the network bandwidth configuration data and the scheduling parameters.

9. A TSN-based industrial real-time data transmission assurance system, used to execute the method according to any one of claims 1 to 8, characterized in that: The system comprises: The TSN network topology deployment module is used to obtain network node information of the industrial site, perform TSN topology deployment and configuration according to the network node information, and obtain an industrial network topology structure that supports the TSN protocol; The system time synchronization module is used to obtain the local system time information of each TSN network node in the industrial network topology, accurately synchronize the local system time information, and obtain a unified system time benchmark for the entire network; A task parameter extraction module is used to obtain historical communication data and service characteristic data of various task flows in the industrial field, perform feature extraction on the historical communication data and service characteristic data, and obtain task parameters of the various task flows; wherein the task parameters include periodicity, maximum delay tolerance, bandwidth requirement, jitter tolerance, and priority level; A load prediction module is used to obtain historical load data of various task flows, time-align the historical load data according to the system time reference, construct a unified time series, input the unified time series into a preset prediction model, and obtain load prediction results for various task flows; The TSN simulation module is used to input the system time base, task parameters, preset scheduling strategy list and load forecast results into the preset TSN digital twin network model, perform simulations of multiple scheduling strategies, and obtain performance indicator data under each scheduling strategy; The scheduling optimization module is used to determine the optimal scheduling strategy based on the performance indicator data and the task parameters, send the optimal scheduling strategy to the TSN switching device, and complete the scheduling control configuration of various task flows in the network.

10. An electronic device, characterized in that: The invention comprises a processor, a memory, and a program or instruction stored in the memory and executable on the processor, wherein the program or instruction, when executed by the processor, implements the steps of the TSN-based industrial real-time data transmission guarantee method according to any one of claims 1 to 8.

Images