5G-driven digital factory AI adaptive production optimization method and system

Abstract

The invention relates to a 5G-driven digital factory AI adaptive production optimization method and system, and relates to the technical field of digital factory optimization, and the method comprises the steps: constructing a 5G-TSN fusion network base layer, and training a semantic understanding and packaging model in combination with a historical industrial control data flow; a deterministic transmission channel is established for the key production process, and a related control instruction stream is packaged into a standardized data unit; monitoring the operation state data of the deterministic transmission channel and the end-to-end transmission performance index of the packaged data unit in real time, inputting the channel parameter dynamic optimization model, and outputting an adjustment strategy; and dynamically reconfiguring the key parameters according to the adjustment strategy, and iteratively executing the whole process by taking the reconfigured transmission performance index as feedback until the transmission of the control instruction stream meets a preset deterministic performance target. The problems that communication protocols of traditional production equipment are disordered, adaptation is difficult, transmission is unstable, and the production efficiency of factories is restricted are solved.

Description

Technical Field

[0001] This application relates to the field of digital factory optimization, and in particular to a 5G-driven AI-adaptive production optimization method and system for digital factories. Background Technology

[0002] With the accelerated upgrading of digital factories to intelligent systems, 5G technology is being applied more and more widely in industrial scenarios, and seamless transparency and transmission stability of communication between devices have become core technical requirements.

[0003] Currently, a large number of existing production equipment use inconsistent communication protocols, while 5G networks are IP-based. Traditional methods require complex gateways for protocol conversion, which is not only cumbersome but also prone to introducing additional latency and fault points, thus hindering factory production efficiency and the large-scale application of 5G technology. Summary of the Invention

[0004] This application provides a 5G-driven AI adaptive production optimization method and system for digital factories, which improves the current situation of messy communication protocols, difficult adaptation and unstable transmission of production equipment, ensures the reliability of industrial control command transmission and improves the production efficiency of digital factories.

[0005] The embodiments of this application disclose the following technical solutions: In a first aspect, embodiments of this application provide a 5G-driven AI-adaptive production optimization method for digital factories, the method comprising: Construct a 5G-TSN converged network base layer that matches the target production area, and train a semantic understanding and encapsulation model for multiple industrial protocols by combining historical industrial control data streams. In the 5G-TSN converged network base layer, deterministic transmission channels are established for key production processes, and based on the semantic understanding and encapsulation model, control command streams of at least one non-IP industrial protocol involved in the key production processes are encapsulated into standardized data units that can be transmitted through the deterministic transmission channels. The system monitors the operational status data of the deterministic transmission channel and the end-to-end transmission performance indicators of the encapsulated data units in real time. It inputs the operational status data and transmission performance indicators into a pre-built dynamic optimization model for channel parameters and outputs adjustment strategies for the key parameters of the current channel. According to the adjustment strategy, the key parameters of the deterministic transmission channel are dynamically reconfigured, and the reconfigured transmission performance indicators are used as feedback to iteratively execute the entire process from monitoring performance indicators to dynamic reconfiguration until the transmission of control command streams for the key production process meets the preset deterministic performance target.

[0006] Secondly, embodiments of this application provide a 5G-driven AI-adaptive production optimization system for digital factories, the system comprising: The network infrastructure construction and model training module is used to build a 5G-TSN converged network base layer that matches the target production area, and to train a semantic understanding and encapsulation model for multiple industrial protocols by combining historical industrial control data streams. The channel establishment and protocol encapsulation module is used to establish a deterministic transmission channel for key production processes in the 5G-TSN converged network base layer, and based on the semantic understanding and encapsulation model, encapsulate the control command stream of at least one non-IP industrial protocol involved in the key production processes into a standardized data unit that can be transmitted through the deterministic transmission channel. The performance monitoring and strategy output module is used to monitor the operating status data of the deterministic transmission channel and the end-to-end transmission performance indicators of the encapsulated data units in real time, input the operating status data and transmission performance indicators into the pre-built channel parameter dynamic optimization model, and output adjustment strategies for the key parameters of the current channel. The parameter reconfiguration and iterative optimization module is used to dynamically reconfigure the key parameters of the deterministic transmission channel according to the adjustment strategy, and use the reconfigured transmission performance index as feedback to iteratively execute the entire process from monitoring performance index to dynamic reconfiguration until the transmission of control command stream of the key production process meets the preset deterministic performance target.

[0007] One or more technical solutions provided in this application have at least the following technical effects or advantages: This application proposes a 5G-driven AI-adaptive production optimization method and system for digital factories. By conducting data acquisition and preprocessing in stages, constructing a dynamic optimization model for channel parameters, generating parameter adjustment strategies, executing dynamic reconfiguration of key parameters, and closed-loop iterative optimization, deterministic and adaptive optimization of the transmission of control command streams for critical production processes is achieved. First, industrial control data of critical production processes is collected, and the operating status data and transmission performance indicators of deterministic transmission channels are acquired simultaneously. Multi-dimensional data preprocessing is completed by combining the flow characteristics and equipment characteristics of historical industrial control data streams. Then, a dynamic optimization model for channel parameters, encompassing protocol adaptation constraints and latency trade-offs, is constructed based on historical data. Multiple rounds of iterative training ensure the model's decision-making accuracy. Subsequently, the real-time collected operating status data and transmission performance indicators are input into the model to generate parameter adjustment strategies including transmission queue priority, bandwidth resources, and time windows. Next, based on the adjustment strategies, the network management interface is invoked, TSN switch parameters are adjusted, and a scheduling table is issued to complete the dynamic reconfiguration of key channel parameters. Finally, using the reconfigured transmission performance indicators as feedback, the entire process from monitoring to reconfiguration is iteratively executed until the transmission of control command streams meets the preset deterministic performance target.

[0008] The technical solution of this application solves the problems of fixed channel parameter configuration, difficulty in adapting to dynamic changes in production processes, imbalance between transmission performance and resource consumption, and lack of closed-loop optimization mechanism in traditional industrial transmission. It avoids excessive transmission latency, increased jitter, or resource waste caused by network load fluctuations and production demand adjustments. It improves the transmission stability, real-time responsiveness, and resource utilization efficiency of key production processes in digital factories, and provides reliable technical support for efficient collaboration and stable operation of factory production. Attached Figure Description

[0009] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0010] Figure 1 A flowchart illustrating the 5G-driven AI adaptive production optimization method for digital factories provided in this application embodiment; Figure 2 A schematic diagram of the structure of the 5G-driven digital factory AI adaptive production optimization system provided in the embodiments of this application.

[0011] The components represented by each number in the attached diagram are explained below: Module 01 for network infrastructure construction and model training, Module 02 for channel establishment and protocol encapsulation, Module 03 for performance monitoring and policy output, and Module 04 for parameter reconfiguration and iterative optimization. Detailed Implementation

[0012] This application provides a 5G-driven AI-adaptive production optimization method and system for digital factories, which addresses the technical problems in existing technologies such as messy communication protocols of production equipment, difficulty in adaptation, and unstable transmission, thereby restricting factory production efficiency.

[0013] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0014] In the description of this application, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the stated features. In the description of this application, "multiple" means two or more, unless otherwise explicitly specified.

[0015] In the description of this application, the term "for example" is used to mean "used as an example, illustration, or description." Any embodiment described as "for example" in this application is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is provided to enable any person skilled in the art to make and use the invention. Details are set forth in the following description for purposes of explanation. It should be understood that those skilled in the art will recognize that the invention can be made without using these specific details. In other instances, well-known structures and processes will not be described in detail to avoid obscuring the description of the invention with unnecessary detail. Therefore, the invention is not intended to be limited to the embodiments shown, but is consistent with the broadest scope of the principles and features disclosed in this application.

[0016] Example 1, as shown in the appendix Figure 1 As shown, this application provides a 5G-driven AI-adaptive production optimization method for digital factories, the method comprising the following steps: S110: Construct a 5G-TSN converged network base layer that matches the target production area, and train a semantic understanding and encapsulation model for multiple industrial protocols by combining historical industrial control data streams. In this embodiment of the application, in the scenario of collaborative operation of multiple types of industrial equipment in a digital factory and coexistence of different communication protocols, in order to build a stable, reliable industrial communication foundation that is compatible with multiple protocols, and at the same time achieve seamless connection between non-IP industrial protocols and IP-based 5G networks, it is necessary to simultaneously complete the construction of a dedicated network foundation and the training of a protocol semantic parsing model, so as to provide dual support for the deterministic transmission and dynamic optimization of subsequent control commands.

[0017] Specifically, based on the physical distribution of industrial equipment within the target production area and the communication relationships between the equipment, 5G base stations and TSN switches are rationally planned and deployed to ensure full signal coverage and complete communication links for the equipment, forming a 5G-TSN converged network foundation layer capable of carrying industrial control command transmission.

[0018] The method provided in this application embodiment includes the following steps for constructing the 5G-TSN converged network base layer: Based on the physical distribution and communication relationship of industrial equipment in the target production area, 5G base stations and TSN switches are deployed to form a 5G-TSN converged network infrastructure layer covering the target production area. Configure the time synchronization mechanism of the 5G-TSN converged network base layer, and divide and pre-configure transmission queues with different priorities and time windows according to the periodicity and timeliness constraints of the key production processes. The key production processes include periodic scheduling processes, event-driven processes, and real-time control processes. Based on the traffic characteristics of the historical industrial control data streams, bandwidth resources and routing strategies are initially allocated to the transmission queues in the 5G-TSN converged network base layer to form an initial deterministic transmission channel resource pool.

[0019] Specifically, the first step is to comprehensively analyze the physical distribution of industrial equipment within the target production area, including the installation location, quantity, and spatial layout of the equipment. At the same time, it is necessary to clarify the communication relationships between the equipment, such as which equipment requires high-frequency interactive control commands and which equipment belongs to the upstream and downstream nodes of the same production chain.

[0020] Furthermore, based on the obtained information, the deployment locations and coverage areas of 5G base stations are rationally planned to ensure that base station signals can cover all industrial equipment without dead zones, avoiding communication interruptions caused by signal blind spots. At the same time, according to the equipment distribution density and communication traffic requirements, a corresponding number of TSN switches are deployed to realize link connections between devices and base stations, and between devices, ultimately forming a 5G-TSN converged network infrastructure layer with complete coverage and smooth links.

[0021] For example, in a large automobile assembly workshop, 5G base stations can be deployed at the four corners and the center of the workshop, and TSN switches can be deployed near each production workstation cluster to ensure that the communication needs of equipment such as welding robots and handling robotic arms are met.

[0022] Furthermore, a time synchronization mechanism is configured for the 5G-TSN converged network base layer, using a mature high-precision synchronization protocol from industrial scenarios to ensure that the clocks of all devices in the network remain consistent, thus avoiding timing errors in the transmission of control commands due to time deviations.

[0023] Simultaneously, based on the periodicity and timeliness constraints of key production processes, transmission queues with different priorities and time windows are divided and pre-configured. Specifically, for periodically scheduled processes, control instructions need to be sent at fixed intervals, requiring high transmission stability; therefore, higher priority and smaller time windows can be allocated to ensure timely transmission of instructions. For event-driven processes, instruction triggering is random, requiring flexible time windows and allocation of medium priority.

[0024] Furthermore, real-time control processes directly impact production safety and accuracy, requiring extremely high latency. They are assigned the highest priority and smallest time window to ensure immediate response to commands. For example, spindle speed control in machine tool processing is a real-time control process, assigned the highest priority, and has a time window granularity of 1ms; material warehousing scheduling commands are event-driven processes, assigned medium priority, and have a time window granularity of 5ms.

[0025] Furthermore, a resource allocation scheme for transmission queues is constructed based on the traffic characteristics of historical industrial control data streams. Specifically, by analyzing historical industrial control data streams, the traffic characteristics such as instruction traffic size, peak periods, and transmission frequency corresponding to different types of production processes are identified. For example, the peak instruction traffic for a certain periodic scheduling process is 10 Mbps, and the transmission period is 100 ms.

[0026] Based on the above characteristics, appropriate bandwidth resources are initially allocated to each transmission queue to ensure stable transmission even during peak traffic periods, avoiding packet loss or increased latency due to insufficient bandwidth. Simultaneously, an optimal routing strategy is planned, selecting the shortest and least disruptive route based on factors such as physical distance between devices and link load, thereby reducing latency and jitter during transmission.

[0027] For example, for groups of devices with high-frequency interactions, dedicated routes can be planned to avoid interference caused by sharing links with other low-priority traffic. Through the above bandwidth allocation and routing planning, an initial deterministic transmission channel resource pool is formed, providing dedicated resource guarantees for the transmission of control commands for different types of production processes, thereby ensuring that each type of command can obtain transmission conditions that meet its requirements.

[0028] Meanwhile, throughout the entire construction process, attention must be paid to the coordination and adaptation of each step. For example, the deployment locations of 5G base stations and TSN switches must match the coverage of subsequent transmission queues, the accuracy of the time synchronization mechanism must meet the timing requirements of high-priority queues, and the allocation of bandwidth resources must be comprehensively considered in conjunction with historical traffic characteristics and queue priorities.

[0029] If high communication latency is found in a certain area, the deployment location of the base station can be adjusted or the number of TSN switches can be increased. If a queue has insufficient bandwidth, the bandwidth allocation scheme can be re-optimized based on the dynamic changes in historical traffic data to ensure that the 5G-TSN converged network infrastructure layer can continuously adapt to the communication needs of the production process and lay the foundation for the efficient transmission of subsequent industrial control commands.

[0030] Furthermore, by analyzing historical industrial control data streams, a semantic understanding and encapsulation model for multiple industrial protocols is constructed. This enables accurate parsing and unified encapsulation of control commands from different protocols, thereby eliminating communication barriers caused by protocol differences and ensuring that non-IP industrial protocol commands can be seamlessly adapted to the IP-based 5G-TSN converged network infrastructure layer.

[0031] The method provided in this application includes the following steps in constructing the semantic understanding and encapsulation model: Based on the historical industrial control data stream, instruction stream samples containing at least one non-IP industrial protocol are collected, and the protocol fields, control semantics, and instruction timing relationships in the instruction stream samples are parsed and annotated to obtain an annotated set of industrial protocol instruction samples. Based on machine learning, the initial network architecture of the semantic understanding and encapsulation model is constructed; The initial network architecture is trained under supervised supervision using the industrial protocol instruction sample set until the verification accuracy converges, thus completing the construction of the semantic understanding and encapsulation model.

[0032] Specifically, the first step is to select and collect instruction stream samples covering different production processes and equipment types from historical industrial control data streams. This ensures that the samples include fixed-cycle instructions for periodic scheduling processes, random-triggered instructions for event-driven processes, and high-priority instructions for real-time control processes. It also takes into account different versions and formats of common non-IP industrial protocols to ensure the comprehensiveness and representativeness of the samples.

[0033] For example, various types of instruction streams, including PLC control protocols and sensor data transmission protocols, are collected. These streams include operation instructions such as equipment start-up and parameter adjustment, as well as response instructions such as status feedback and fault alarms, forming a sufficiently large instruction stream sample pool.

[0034] Furthermore, the collected instruction stream samples are analyzed and labeled. Specifically, the field structure of the instruction stream is disassembled frame by frame, clarifying the encoding format, length, and functional definition of each field. For example, key parts such as instruction headers, data segments, and check bits are distinguished, and the corresponding equipment control functions of the fields are labeled, such as spindle speed setting, material conveying start and stop, and other control semantics.

[0035] Simultaneously, the temporal relationships between instructions are analyzed, and the temporal relationships of the complete chain such as "instruction sending - execution response - status feedback" are labeled. For example, after the device start instruction is sent, the time interval and associated identifier of the subsequent corresponding operation status feedback instructions are formed, forming a set of industrial protocol instruction samples with triple annotation of field information, semantic connotation and temporal relationship, providing labeled data for model training.

[0036] Furthermore, an initial network architecture for semantic understanding and encapsulation models is constructed based on machine learning. Combining the serialization characteristics and semantic relevance requirements of industrial protocol instructions, Transformer is selected as the basic network architecture. Its self-attention mechanism can effectively capture the dependencies and temporal features between instruction fields, adapting to the complex requirements of multi-protocol parsing.

[0037] Specifically, during the architecture construction process, network parameters are set reasonably according to the size of the sample set and the complexity of the protocol types: if the sample size is 50,000 to 80,000 and covers 6 to 8 non-IP industrial protocols, a network structure containing 6 layers of encoders and 4 layers of decoders can be constructed. The encoder input layer receives a fixed-length instruction sequence, and each encoder layer is equipped with 8 attention heads and uses the ReLU activation function.

[0038] In addition, the decoder output layer uses the Softmax function to output the protocol field parsing results and semantic identifiers. If the sample size is less than 30,000 and the protocol types are less than 4, it can be simplified to 4 encoder layers and 3 decoder layers, reducing the number of attention heads to 6 and reducing the network parameter size to avoid overfitting.

[0039] Furthermore, a labeled set of industrial protocol instructions was used to supervise the training of the initial network architecture. The sample set was divided into a training set, a validation set, and a test set in a 7:2:1 ratio. The training set was used for iterative updates of model parameters, the validation set was used to monitor performance changes during training, and the test set was used for final model performance evaluation.

[0040] During training, instruction stream samples are input into the model, and the model outputs field parsing results, semantic recognition results, and temporal relationship judgment results. These are compared with the sample annotation information. The mean squared error (MSE) is used to quantify the field parsing bias, and the cross-entropy loss is used to quantify the semantic recognition error. These are combined to form the total loss function. Based on the Adam optimization algorithm, network weights, attention coefficients, and other parameters are adjusted in reverse according to the total loss value to continuously optimize model performance.

[0041] During training, periodic validation and parameter adjustments are necessary to ensure stable model convergence. For example, after every 20 training rounds, the model's parsing accuracy and semantic recognition accuracy should be tested using a validation set. If the accuracy does not improve or even decreases after 5 consecutive validation rounds, training parameters should be adjusted promptly: if the recognition accuracy stagnates around 85%, the learning rate can be reduced from 0.001 to 0.0005; if overfitting occurs (training set accuracy 98%, validation set accuracy 82%), a Dropout layer can be added (dropout rate set to 0.2) or 2000 more diverse samples can be added. If the parsing accuracy for a certain protocol is consistently below 80%, instruction samples for that protocol should be added to enhance the model's adaptability to less common protocols.

[0042] Furthermore, when the overall accuracy on the validation set remains stable above 95% for 10 consecutive rounds and the accuracy on the test set is not lower than 93%, the model training is considered to have converged, and the construction of the semantic understanding and encapsulation model is completed.

[0043] Ultimately, the constructed semantic understanding and encapsulation model can accurately parse the field structure and control semantics of different non-IP industrial protocols, and capture the timing correlation of instructions. It can quickly transform heterogeneous protocol instructions into standardized data that meets network transmission requirements, providing reliable protocol adaptation support for deterministic transmission in 5G-TSN converged networks.

[0044] S120: In the 5G-TSN converged network base layer, a deterministic transmission channel is established for key production processes, and based on the semantic understanding and encapsulation model, the control command stream of at least one non-IP industrial protocol involved in the key production processes is encapsulated into a standardized data unit that can be transmitted through the deterministic transmission channel. In this embodiment of the application, in order to solve the problem of adapting to multiple types of non-IP industrial protocols and ensure the determinism and stability of the transmission of control commands for key production processes, it is necessary to break down the barriers between protocol differences and network transmission by establishing a dedicated transmission channel and standardizing the encapsulation of command data, so as to support efficient collaboration and precise control of the production process.

[0045] First, from the initial deterministic transmission channel resource pool, at least one transmission queue with corresponding priority and time window is allocated to the key production process. Based on the traffic characteristics of historical industrial control data flow, initial bandwidth resources and initial routing strategies are configured for the transmission queue to complete the construction of the deterministic transmission channel, so as to ensure that the channel resources are accurately matched with the timeliness and periodicity requirements of the production process.

[0046] Furthermore, based on the obtained semantic understanding and encapsulation model, the non-IP industrial protocol control command flow involved in key production processes is parsed in real time, and the protocol command structure, field semantics and control logic are extracted in depth to eliminate semantic barriers between different protocols and provide a unified understanding basis for subsequent data encapsulation.

[0047] Finally, based on the transmission requirements of the deterministic transmission channel, the semantically understood control commands are adapted, a timestamp is generated based on the time synchronization mechanism, a priority label is added to the transmission queue, a unique channel identifier is used as the channel identifier, and the control commands are segmented or combined to form a data payload according to the data unit length constraint. The timestamp, channel identifier, priority label and data payload are integrated and encapsulated according to a preset format, and finally a standardized data unit that can be stably transmitted through the deterministic transmission channel is generated.

[0048] Step S120 in the method provided in this application embodiment includes: From the initial deterministic transmission channel resource pool, allocate at least one transmission queue with corresponding priority and time window for the key production process, and configure initial bandwidth resources and initial routing strategies for the transmission queue based on the traffic characteristics of the historical industrial control data stream, and establish a corresponding deterministic transmission channel. Using the semantic understanding and encapsulation model, the non-IP industrial protocol control command flow involved in the key production process is parsed and semantically understood in real time, and the protocol command structure, field semantics and control logic are extracted. Based on the transmission requirements of the deterministic transmission channel, the semantically understood control instructions are adapted and encapsulated to generate standardized data units with timestamps, channel identifiers, and priority tags.

[0049] Specifically, the first step is to select transmission queues suitable for critical production processes from the initial deterministic transmission channel resource pool. Priority requirements are then defined based on the type of critical production process: real-time control processes are assigned the highest priority, periodically scheduled processes are assigned a higher priority, and event-driven processes are assigned a medium priority. Simultaneously, corresponding time windows are matched to ensure that time-sensitive instructions receive priority transmission resources.

[0050] Furthermore, by combining the traffic characteristics of historical industrial control data streams, including the average size of command traffic, peak periods, and transmission frequency, initial bandwidth resources are configured for the allocated transmission queues. The bandwidth capacity needs to reserve a certain redundancy to cope with traffic peaks, so as to avoid packet loss or latency surges caused by insufficient bandwidth.

[0051] At the same time, an initial routing strategy is planned, taking into account the physical distance between devices, the link load balancing status, and interference factors. The route with the shortest transmission path and the highest stability is selected to reduce link loss and latency jitter during transmission, and finally establish a dedicated deterministic transmission channel to ensure the basic stability of command transmission.

[0052] For example, the spindle speed control of a machine tool in an automotive parts processing workshop is a real-time control process. It is assigned the highest priority and a 1ms granularity time window. Based on its historical command traffic characteristics of an average of 2Mbps and a peak of 5Mbps, an initial bandwidth resource of 8Mbps is configured (with 60% redundancy reserved). A dedicated route is planned from the machine tool controller to the nearest TSN switch to directly connect to the 5G base station. This avoids sharing the link with event-driven processes such as material handling (medium priority, 5ms time window, 3Mbps bandwidth), ensuring that the speed control command transmission is without delay or jitter.

[0053] Furthermore, the trained semantic understanding and encapsulation model is used to process non-IP industrial protocol control command streams generated by key production processes in real time. Specifically, the command stream is continuously input into the model, which, through its pre-learned multi-protocol parsing capabilities, deconstructs the field structure of the command frame by frame, clarifies the encoding format, length, functional identifier, and relationships between fields, and extracts the semantic meaning of each field.

[0054] For example, different control semantics are distinguished, such as device start-stop control, operating parameter adjustment, and fault status feedback. At the same time, the timing logic of command sending and response is sorted out, and the control logic of the complete chain such as "command initiation - execution confirmation - status feedback" is clarified, so as to provide semantic support for subsequent adaptation and encapsulation.

[0055] Furthermore, based on the transmission requirements of the deterministic transmission channel, the semantically understood control commands are adapted and encapsulated to generate standardized data units with timestamps, channel identifiers, and priority tags, thereby eliminating transmission barriers caused by heterogeneous protocol differences and ensuring that control commands can be transmitted stably and efficiently in the channel.

[0056] The method provided in this application embodiment adapts and encapsulates semantically understood control instructions according to the transmission requirements of the deterministic transmission channel, generating standardized data units with timestamps, channel identifiers, and priority tags, including: The control command is provided with a transmission time reference according to the time synchronization mechanism, and the timestamp is generated based on the transmission time reference. Based on the priority and time window of the transmission queue allocated to the critical production process, the priority label is determined and marked, and the unique identifier of the deterministic transmission channel is used as the channel identifier. Based on the data unit length constraints determined by the flow characteristics of the historical industrial control data stream, the control commands are segmented or combined to form a data payload that meets the transmission requirements. The timestamp, channel identifier, priority tag, and data payload are combined and encapsulated according to a preset format to generate a standardized data unit that can be transmitted through a deterministic transmission channel.

[0057] Specifically, based on the high-precision time synchronization mechanism already configured in the 5G-TSN converged network base layer, the reference clock signal at the moment the control command is sent is obtained. This reference clock must be consistent with the clocks of all devices in the network to ensure the uniformity of timing data.

[0058] Furthermore, based on the reference clock signal, a timestamp accurate to the microsecond level is generated. The timestamp format adopts standard UTC time encoding, containing year, month, day, hour, minute, second, and microsecond information. For example, if a control command is sent at 10:05:30.124789 microseconds, the corresponding timestamp is "2025-08-20T10:05:30.124789Z". This timestamp accurately records the time of command transmission, providing a reliable time basis for delay calculation and timing calibration in subsequent transmission performance monitoring.

[0059] Furthermore, based on the priority level and time window parameters of the transmission queues allocated to critical production processes, corresponding priority tags are determined. Specifically, priority tags adopt a numerical encoding format, for example, the highest priority is labeled as 001, a higher priority as 002, and a medium priority as 003, ensuring that the tags completely correspond to the priority configuration of the transmission queues, so that network devices can quickly identify and prioritize high-priority instructions during transmission.

[0060] Meanwhile, the unique identifier of the deterministic transmission channel is extracted as the channel identifier. This identifier is a globally unique code assigned during network deployment, such as "TSN-CH-202508-007". The channel identifier can clearly identify the exclusive transmission path corresponding to the instruction, avoid path confusion or mistransmission during transmission, and ensure that the instruction can be accurately routed to the target device.

[0061] Furthermore, data unit length constraints are determined based on the flow characteristics of historical industrial control data streams. Specifically, by analyzing the length distribution, transmission frequency, and peak flow of different types of control commands in historical data, and combining this with the maximum frame length limit of the deterministic transmission channel, a data unit length threshold is set, for example, the length range of a single data unit is 64 bytes to 1500 bytes.

[0062] For control commands exceeding the threshold in length, they are segmented according to the principle of semantic integrity to ensure that each segment contains complete field information and semantic fragments. Each segment is also labeled with a sequential number to facilitate reassembly at the receiving end. Furthermore, for multiple related control commands whose length is much lower than the threshold, they are combined without compromising semantic logic, integrating multiple commands into a single data unit to reduce transmission overhead.

[0063] For example, a device parameter configuration instruction is 2000 bytes long, exceeding the 1500-byte threshold. It is semantically split into two segments: the first segment is 1500 bytes and the second segment is 500 bytes, labeled with sequence numbers 01 and 02 respectively. Two related status query instructions are both 80 bytes long, combined into a 160-byte data payload to ensure compliance with length constraints.

[0064] Finally, the generated timestamp, channel identifier, priority label, and processed data stream are combined and encapsulated according to a preset unified format. Specifically, the preset format adopts a "header-payload-tail" structure. The header contains a timestamp (48 bytes), a channel identifier (32 bytes), a priority label (3 bytes), and a data length identifier (2 bytes). The payload is segmented or combined control command data, and the tail is a checksum field (4 bytes), calculated using the CRC32 algorithm, used to verify whether errors occurred during data transmission.

[0065] In addition, the encapsulation process strictly follows byte alignment rules to ensure that the data format conforms to the transmission protocol requirements of the 5G-TSN converged network. For example, the header fields are arranged in a fixed order, the payload data is stored in binary format, and the check field is calculated using the CRC32 algorithm.

[0066] Ultimately, through the encapsulation process described above, standardized data units are generated, enabling control commands from different non-IP industrial protocols to be converted into transmission data in a unified format. This achieves seamless adaptation with deterministic transmission channels, providing a standardized data carrier for subsequent performance monitoring and dynamic optimization, and ensuring efficient and stable transmission of commands in the network.

[0067] S130: Monitor the operating status data of the deterministic transmission channel and the end-to-end transmission performance indicators of the encapsulated data units in real time, input the operating status data and transmission performance indicators into the pre-built channel parameter dynamic optimization model, and output the adjustment strategy for the current key parameters of the channel. In this embodiment of the application, in order to keep abreast of the operation status of the transmission channel and the quality of data transmission, and to provide an accurate basis for optimizing the channel parameters, it is necessary to generate a scientific adjustment strategy through comprehensive monitoring and intelligent analysis to ensure the stability and determinism of the transmission of control commands for key production processes.

[0068] First, real-time data on the operational status of deterministic transmission channels is collected, including queue occupancy rates of transmission queues allocated for critical production processes, time deviations of time synchronization mechanisms, and real-time status of relevant links in the 5G-TSN converged network infrastructure layer, comprehensively capturing core status information during channel operation.

[0069] Meanwhile, the end-to-end transmission performance indicators of the encapsulated data units are collected in real time, including end-to-end latency, latency jitter, and packet loss rate, which accurately reflect the actual performance of the data during transmission.

[0070] Subsequently, the collected operational status data and transmission performance indicators are input into a pre-built dynamic optimization model for channel parameters. This model will combine the equipment characteristics of industrial equipment involved in key production processes with the unified constraints of protocol configuration based on semantic understanding and encapsulation models to conduct collaborative analysis.

[0071] In the collaborative analysis process, the additional latency caused by protocol adaptation and encapsulation is quantified as the optimization cost, while the latency reduction brought about by the optimized configuration of the transmission channel is quantified as the optimization value, thereby clearly defining the costs and benefits in the optimization process.

[0072] Finally, with minimizing overall latency and meeting preset deterministic performance targets as the optimization direction, an adjustment strategy is generated based on the trade-off between optimization costs and optimization value. This strategy includes reallocating transmission queue priorities, readjusting initial bandwidth resources, or replanning time windows, providing clear guidance for subsequent dynamic reconfiguration of channel parameters.

[0073] Step S130 in the method provided in this application embodiment includes: Real-time acquisition of operational status data of the deterministic transmission channel, including queue occupancy rate of the transmission queue allocated to the key production process, time deviation of the time synchronization mechanism, and real-time status of relevant links in the 5G-TSN converged network infrastructure layer. Real-time acquisition of end-to-end transmission performance indicators of the encapsulated data units, including end-to-end latency, latency jitter, and packet loss rate; The collected operating status data and transmission performance indicators are input into the pre-constructed channel parameter dynamic optimization model, wherein the channel parameter dynamic optimization model combines the equipment characteristics of the industrial equipment involved in the key production process with the protocol configuration uniformity constraints based on the semantic understanding and encapsulation model to perform collaborative analysis; In the aforementioned collaborative analysis, the additional latency caused by protocol adaptation and encapsulation is quantified as the optimization cost, and the latency reduction brought about by optimized transmission channel configuration is quantified as the optimization value. Based on the optimization cost and optimization value, with the optimization direction of minimizing overall latency and meeting the preset deterministic performance target, an adjustment strategy for the current channel key parameters is output. The adjustment strategy includes at least one of the following: reallocation of transmission queue priority, readjustment of initial bandwidth resources, or replanning of time window.

[0074] In this embodiment of the application, in order to grasp the operational status and data transmission quality of the deterministic transmission channel in real time and provide a comprehensive decision-making basis for the dynamic optimization of channel parameters, it is necessary to generate a scientific and feasible parameter adjustment strategy through multi-dimensional data collection, intelligent model analysis and cost-value balancing, so as to continuously adapt to the dynamic changes in the production process of the digital factory.

[0075] Specifically, the first step is to conduct real-time data acquisition, collecting operational status data of the deterministic transmission channel. The queue occupancy rate directly reflects the queue's load capacity; for example, when the queue occupancy rate consistently exceeds 75%, it may lead to extended data transmission queuing times.

[0076] Meanwhile, the time deviation of the time synchronization mechanism is related to the timing consistency of command transmission. For example, a deviation exceeding 5 microseconds may affect the accuracy of multi-device collaborative control. In addition, the real-time status of relevant links in the 5G-TSN converged network infrastructure layer includes link connectivity, signal strength, and interference conditions. For example, if the signal strength of a link is attenuated by more than 30% due to external interference, it may cause data transmission instability.

[0077] Data acquisition requires high-frequency sampling, and the sampling interval is set according to the response sensitivity requirements of the production process. For example, for real-time control processes, the sampling interval can be set to the millisecond level to ensure that dynamic changes in status data can be captured in a timely manner.

[0078] On the other hand, it involves real-time acquisition of end-to-end transmission performance metrics for the encapsulated data units. End-to-end latency is a metric for measuring transmission efficiency; for example, periodic scheduling processes typically require end-to-end latency to be no more than 20 milliseconds to ensure timely execution of scheduling instructions. Latency jitter reflects the degree of fluctuation in transmission latency; for example, jitter exceeding 3 milliseconds can disrupt the execution rhythm of control instructions, affecting the stability of the production process.

[0079] Furthermore, packet loss rate is related to the integrity of data transmission. For example, a packet loss rate exceeding 0.1% may result in missing control commands, leading to production process interruptions or malfunctions. Simultaneously, performance indicators need to be continuously monitored and recorded during the data acquisition process to ensure data continuity and accuracy, providing reliable support for subsequent analysis.

[0080] After data collection is completed, the above-mentioned operational status data and transmission performance indicators are input into the pre-built channel parameter dynamic optimization model. This model combines the equipment characteristics of industrial equipment involved in key production processes with the constraints of protocol configuration uniformity to conduct collaborative analysis, accurately quantify the optimization cost and value, and output channel parameter adjustment strategies adapted to the current network status.

[0081] Among them, the channel parameter dynamic optimization model is an intelligent optimization model that is trained and generated based on historical industrial control data streams. It can integrate multi-dimensional data to perform time delay trade-off analysis, so as to ensure that it can stably output scientific strategies that meet deterministic performance targets based on real-time input data and guide parameter configuration, thus ensuring that the performance of the transmission channel continuously adapts to production needs.

[0082] Specifically, based on historical industrial control data streams, we first extensively collect historical operating status data and corresponding historical transmission performance indicators of deterministic transmission channels under different protocol configurations and load scenarios, covering data samples under various operating conditions such as high load, low load, and normal load, to form a comprehensive training sample set.

[0083] Furthermore, in the training sample set, the additional latency caused by protocol adaptation and encapsulation is accurately labeled as the optimization cost. For example, after adaptation and encapsulation, a certain non-IP protocol generates an average additional latency of 1.2 microseconds. At the same time, the latency reduction achieved by adjusting key channel parameters is labeled as the optimization value. For example, increasing the bandwidth from 100Mbps to 200Mbps reduces the transmission latency by 4 microseconds.

[0084] Furthermore, by combining the equipment characteristics of industrial equipment involved in key production processes and the unified constraints of protocol configuration based on semantic understanding and encapsulation models, an initial optimization model is constructed with minimizing overall latency as the core optimization objective, ensuring that the optimization direction of the model is consistent with actual production needs.

[0085] Finally, the constructed training sample set is used to iteratively supervise the training of the initial optimization model. By adjusting the model's weight parameters, activation functions, etc., the prediction accuracy and decision-making scientificity of the model are continuously improved until the model can stably output channel key parameter adjustment strategies that meet the requirements of deterministic performance targets based on the input real-time running status data and transmission performance indicators, thus completing the construction of the channel parameter dynamic optimization model.

[0086] Furthermore, after receiving the collected data, the model will conduct collaborative analysis by combining the equipment characteristics and protocol configuration uniformity constraints of the industrial equipment involved in key production processes. Among these, the equipment characteristics of industrial equipment need to be fully considered. For example, some types of processing equipment have low communication rates and relatively stable bandwidth requirements, and the adjustment strategy needs to avoid frequent bandwidth fluctuations affecting its operation.

[0087] Furthermore, the uniformity constraint of protocol configuration requires that adjustments to the strategy must not disrupt the established adaptation relationships between different protocols, ensuring the compatibility of multi-protocol collaborative transmission. During the collaborative analysis process, standardized quantitative methods will be employed to quantify the additional latency caused by protocol adaptation and encapsulation as optimization costs, while simultaneously quantifying the latency reduction resulting from optimized transmission channel configuration as optimization value, thus clearly defining the costs and benefits of optimization actions.

[0088] Specifically, by calculating the data processing time and format conversion time during the protocol encapsulation process, the specific value of the optimization cost is obtained; by comparing the transmission latency data before and after optimization, the quantitative result of the optimization value is determined.

[0089] For example, when a control command of a non-IP industrial protocol is encapsulated, data parsing takes 0.8 microseconds, format conversion and adaptation take 0.5 microseconds, and other auxiliary operations take 0.2 microseconds. The comprehensive calculation shows that the optimization cost corresponding to the encapsulation of this protocol is 1.5 microseconds. Before optimization, the end-to-end latency of this command through the transmission channel is 28 microseconds. After adopting the strategy of adjusting bandwidth resources and queue priority, the end-to-end latency is reduced to 22 microseconds. The optimization value is calculated to be 6 microseconds based on the latency difference.

[0090] Finally, the model takes minimizing the overall latency and meeting the preset deterministic performance target as the core optimization direction. Based on the trade-off between optimization cost and optimization value, it outputs targeted channel key parameter adjustment strategies.

[0091] For example, when the optimization value is significantly higher than the optimization cost, and the end-to-end latency does not reach the preset target, an adjustment strategy to increase the priority of the transmission queue can be output, so that the instruction data of the critical production process can obtain a higher transmission priority.

[0092] In addition, when excessive latency jitter is detected and the link bandwidth utilization is low, an adjustment strategy to increase the initial bandwidth resources can be output to improve transmission stability by increasing the bandwidth reservation value.

[0093] In addition, when unreasonable time window allocation leads to transmission congestion in some periods, an adjustment strategy for rescheduling the time window can be output, and the transmission period can be re-divided based on the time synchronization mechanism to avoid conflicts in the transmission of instruction data from different processes.

[0094] Meanwhile, the above adjustment strategies need to have clear execution paths and parameter thresholds to provide clear and operable guidance for the dynamic reconfiguration of key parameters of subsequent deterministic transmission channels, ensuring continuous optimization of transmission performance and always meeting the transmission needs of critical production processes.

[0095] S140: According to the adjustment strategy, the key parameters of the deterministic transmission channel are dynamically reconfigured, and the reconfigured transmission performance index is used as feedback to iteratively execute the entire process from monitoring performance index to dynamic reconfiguration until the transmission of control command stream of the key production process meets the preset deterministic performance target.

[0096] In this embodiment of the application, in order to transform the channel parameter adjustment strategy into actual transmission performance improvement and continuously adapt to the dynamic changes and transmission requirements of the production process, it is necessary to ensure that the deterministic transmission channel always maintains the optimal operating state through parameter reconfiguration operation and closed-loop iterative optimization mechanism, so as to ensure the efficient collaboration and stable operation of factory production.

[0097] Specifically, based on the obtained adjustment strategy, the key parameters of the deterministic transmission channel are dynamically reconfigured to quickly adapt to the transmission needs of critical production processes and changes in network operating status, thereby optimizing the determinism and stability of channel transmission.

[0098] The method provided in this application embodiment dynamically reconfigures the key parameters of the deterministic transmission channel according to the adjustment strategy, including: According to the reallocation suggestion for transmission queue priority in the adjustment strategy, the priority label of the transmission queue bound to the deterministic transmission channel is updated by calling the network management interface of the 5G-TSN converged network base layer. Based on the readjustment recommendations for initial bandwidth resources in the adjustment strategy, the bandwidth resource reservation value corresponding to the transmission queue is updated by adjusting the traffic shaping parameters of the transmission queue in the TSN switch. Based on the time window replanning recommendations in the adjustment strategy, the available time window boundaries of the transmission queue are adjusted by recalculating and issuing the scheduling table based on the time synchronization mechanism.

[0099] Specifically, the first step is to perform a reallocation operation of transmission queue priorities. The priority of the transmission queue directly determines the transmission priority of the control command stream in network resource contention. Based on the specific recommendations of the adjustment strategy, the priority of the transmission queue needs to be updated by directly accessing the configuration management module of the transmission queue through the network management interface of the 5G-TSN converged network infrastructure layer.

[0100] The priority allocation should be combined with the type and timeliness requirements of the key production process. For example, real-time control processes have extremely high requirements for real-time transmission, so their corresponding transmission queue priority can be raised from level two to level one to ensure that transmission resources can still be obtained first when the network load is high.

[0101] Furthermore, periodic scheduling processes can maintain a three-level priority; event-driven processes can be adjusted to a second-level or first-level priority based on their urgency. During the update process, it is necessary to ensure that the modification of priority labels is performed within the authorization scope of the network management interface and is synchronously updated to the collaborative configuration of the 5G base station and TSN switch to avoid transmission conflicts caused by configuration asynchrony.

[0102] For example, when the adjustment strategy suggests raising the priority of the transmission queue of real-time control processes from level two to level one, a priority update instruction is sent through the network management interface to change the priority label of the queue from level two priority to level one priority, and synchronized to the queue scheduling module of the 5G base station and the priority processing unit of the TSN switch to ensure that the instruction can be scheduled with priority during transmission.

[0103] Furthermore, based on the readjustment recommendations for initial bandwidth resources in the adjustment strategy, the traffic shaping parameters of the transmission queue in the TSN switch are adjusted, and the bandwidth resource reservation value corresponding to the transmission queue is updated to accurately match the real-time bandwidth requirements of critical production processes and alleviate transmission latency and jitter issues.

[0104] The method provided in this application embodiment updates the bandwidth resource reservation value corresponding to the transmission queue by adjusting the traffic shaping parameters of the transmission queue in the TSN switch according to the readjustment suggestion for the initial bandwidth resources in the adjustment strategy, including: Based on the flow characteristics of the historical industrial control data stream, the baseline bandwidth requirements corresponding to the key production process are determined. Obtain the end-to-end transmission performance indicators of the encapsulated data units collected in real time; The end-to-end latency, latency jitter, and packet loss rate are compared with the preset deterministic performance target to calculate the performance deviation. Based on the optimization costs and values ​​derived from the collaborative analysis, the latency trade-offs under the current network configuration are evaluated, and the baseline bandwidth requirement is proportionally adjusted based on the performance deviation to calculate the updated bandwidth resource reservation value. The updated bandwidth resource reservation value is sent to the TSN switch via configuration commands, and the traffic shaping parameters corresponding to the transmission queue are adjusted to complete the dynamic reconfiguration of bandwidth resources.

[0105] Specifically, the baseline bandwidth requirements for key production processes are first determined based on the flow characteristics of historical industrial control data streams. These flow characteristics are obtained through long-term statistical analysis, including key information such as peak flow, average flow, flow fluctuation frequency, and duration under different time periods and operating conditions.

[0106] For example, the control command sending of periodic scheduling processes has a fixed periodicity and small traffic fluctuations. By analyzing the historical data of the past 3 months, its average traffic is stable at 80Mbps and the peak does not exceed 100Mbps. Therefore, the baseline bandwidth requirement of this process can be set to 100Mbps, which not only meets the regular transmission requirements, but also reserves some space for sudden traffic.

[0107] In addition, the traffic of event-driven processes is random, with peak values ​​potentially reaching 150Mbps and an average of 60Mbps. The baseline bandwidth requirement can be set to 120Mbps to balance resource consumption and transmission assurance.

[0108] At the same time, when determining the baseline bandwidth requirement, the core characteristics of the production process must be fully considered to ensure that the baseline value is not lower than the minimum transmission requirement, while avoiding excessive reservation that would lead to resource waste.

[0109] Furthermore, the end-to-end transmission performance metrics of the encapsulated data units collected in real time are obtained. These performance metrics are the core basis for judging whether the current bandwidth configuration is reasonable, including end-to-end latency, latency jitter, and packet loss rate.

[0110] Specifically, data acquisition should employ high-frequency, high-precision acquisition methods, with the acquisition interval set according to the response sensitivity requirements of the production process. For example, for real-time control processes, the acquisition interval can be set to 1 millisecond to ensure timely capture of subtle changes in transmission performance. During the acquisition process, data must be verified in real time to eliminate abnormal data caused by equipment failure or signal interference, ensuring the authenticity and reliability of performance indicators.

[0111] For example, by verifying the consistency of three consecutive data collections, if the difference between the latency value of a certain collection and the previous two exceeds 5 milliseconds, it is determined to be abnormal data and removed to avoid interfering with subsequent deviation calculations.

[0112] Furthermore, the end-to-end latency, latency jitter, and packet loss rate are compared with preset deterministic performance targets to calculate the performance deviation. The preset deterministic performance targets need to be set comprehensively based on the actual needs of the production process and network capabilities; the target values ​​will differ for different types of production processes.

[0113] For example, the preset end-to-end latency target for real-time control processes is no more than 15 milliseconds, the latency jitter target is no more than 2 milliseconds, and the packet loss rate target is no more than 0.05%; the preset end-to-end latency target for periodic scheduling processes is no more than 30 milliseconds, the latency jitter target is no more than 5 milliseconds, and the packet loss rate target is no more than 0.1%. Performance deviation is calculated using the difference between the real-time monitored value and the preset target value. If the real-time end-to-end latency is 18 milliseconds and the preset target is 15 milliseconds, then the latency deviation is 3 milliseconds; if the real-time latency jitter is 3 milliseconds and the preset target is 2 milliseconds, then the jitter deviation is 1 millisecond; if the real-time packet loss rate is 0.08% and the preset target is 0.05%, then the packet loss rate deviation is 0.03%. By quantifying the deviation, the gap between the current bandwidth configuration and the ideal state is clearly defined.

[0114] Furthermore, by combining the optimization costs and values ​​derived from the collaborative analysis, the latency trade-offs under the current network configuration are assessed, and the baseline bandwidth requirements are proportionally adjusted based on performance deviations. This leads to the calculation of updated bandwidth resource reservation values, enabling precise adaptation and efficient utilization of bandwidth resources, and balancing transmission performance improvement with resource consumption costs.

[0115] The method provided in this application embodiment combines the optimization cost and optimization value obtained from collaborative analysis to evaluate the latency trade-off state under the current network configuration, and adjusts the baseline bandwidth requirement proportionally based on the performance deviation, including: The optimization value is compared with the optimization cost, and it is determined whether the current network configuration is in an effective optimization range based on whether the optimization value exceeds the optimization cost. Obtain the values ​​and signs of each deviation index in the performance deviation, and identify the deviation type and degree of deviation that dominate the performance deviation; Based on the determination result of the effective optimization interval and the degree of deviation from the dominant deviation type, the bandwidth adjustment coefficient is determined; Based on the bandwidth adjustment coefficient, the baseline bandwidth requirement is multiplicatively adjusted, wherein the adjustment direction is determined by the overall sign of the performance deviation, and the adjustment magnitude is jointly constrained by the bandwidth adjustment coefficient and the bandwidth safety margin defined in the flow characteristics of the historical industrial control data stream.

[0116] Specifically, the optimization value is first compared with the optimization cost to determine whether the current network configuration is within an effective optimization range. The optimization value refers to the expected improvement in transmission performance that can be achieved through bandwidth adjustment, mainly reflected in the reduction of end-to-end latency. For example, increasing bandwidth can reduce transmission latency from 25 milliseconds to 20 milliseconds, corresponding to an optimization value of 5 milliseconds.

[0117] In addition, optimization cost refers to the additional costs incurred during bandwidth adjustment, including additional latency caused by protocol adaptation and encapsulation, and compression of other process resources due to additional bandwidth occupation. For example, after adjusting the bandwidth, the additional latency of protocol adaptation increases by 1 microsecond, while occupying 10Mbps of bandwidth resources in other processes. These are all quantified as optimization cost.

[0118] Specifically, if the optimization value significantly exceeds the optimization cost, it indicates that there is room for optimization in the current network configuration, and it is in an effective optimization range, so bandwidth can be increased. If the optimization value is equal to or lower than the optimization cost, it indicates that bandwidth adjustment cannot bring positive benefits, and it is in an ineffective optimization range, so the existing bandwidth should be maintained or appropriately reduced to avoid wasting resources.

[0119] For example, when the optimization value is 5 milliseconds and the optimization cost is 2 milliseconds, it is determined to be in the effective optimization range; when the optimization value is 1 millisecond and the optimization cost is 1.5 milliseconds, it is determined to be in the invalid optimization range.

[0120] Furthermore, the values ​​and signs of each deviation indicator in the performance deviation are obtained to identify the dominant deviation type and degree of deviation. Performance deviations include end-to-end latency deviation, latency jitter deviation, and packet loss rate deviation. The values ​​and signs of each indicator directly reflect the shortcomings in transmission performance. For example, if the preset end-to-end latency target is 20 milliseconds, the real-time monitoring value is 28 milliseconds, and the latency deviation is +8 milliseconds; if the preset latency jitter target is 3 milliseconds, the real-time monitoring value is 5 milliseconds, and the jitter deviation is +2 milliseconds; if the preset packet loss rate target is 0.1%, the real-time monitoring value is 0.2%, and the packet loss rate deviation is +0.1%.

[0121] By comparing the absolute values ​​of each deviation indicator, the type of dominant performance deviation can be identified. In the example above, the absolute value of the latency deviation is the largest, therefore it is determined to be the dominant deviation type, and the degree of deviation is relatively large. If the packet loss rate deviation is +0.5% in a certain scenario, while the latency deviation and jitter deviation are both less than +1, then the packet loss rate deviation is determined to be the dominant deviation type, and the degree of deviation is moderate.

[0122] Furthermore, a positive sign indicates that the performance has not met the preset target, while a negative sign indicates that the performance is better than the preset target. Identifying the dominant deviation type provides a core basis for the direction of subsequent bandwidth adjustments.

[0123] Furthermore, based on the judgment results of the effective optimization interval and the degree of deviation from the dominant deviation type, the bandwidth adjustment coefficient is determined. The magnitude of the bandwidth adjustment coefficient directly determines the extent of bandwidth adjustment and must be set comprehensively based on both types of judgment results to ensure that the bandwidth adjustment coefficient both meets the performance improvement requirements and does not exceed a reasonable range.

[0124] Specifically, if the current area is within the effective optimization range and the dominant deviation type deviates significantly, a larger bandwidth adjustment coefficient can be set, such as 1.4; if the area is within the effective optimization range but the dominant deviation deviates moderately, the bandwidth adjustment coefficient can be set to 1.2; if the area is within the effective optimization range and the dominant deviation deviates slightly, the bandwidth adjustment coefficient can be set to 1.1.

[0125] Conversely, if the system is in an ineffective optimization zone, but some metrics are slightly below the preset target (deviation sign is negative), a bandwidth adjustment coefficient less than 1 can be set to release redundant resources, such as 0.9. For example, if the effective optimization zone has a significantly excessive dominant latency deviation, the bandwidth adjustment coefficient is set to 1.4; if the effective optimization zone has a moderately excessive dominant packet loss rate deviation, the bandwidth adjustment coefficient is set to 1.2; and if the ineffective optimization zone has redundant performance in some metrics, the bandwidth adjustment coefficient is set to 0.9.

[0126] Meanwhile, the setting of the bandwidth adjustment coefficient should take into account historical adjustment data and the total amount of network resources to avoid network load imbalance caused by an excessively large coefficient.

[0127] Finally, based on the determined bandwidth adjustment coefficient, the baseline bandwidth requirement is multiplicatively adjusted. The direction of adjustment is determined by the overall sign of the performance deviation, and the adjustment magnitude is jointly constrained by the bandwidth adjustment coefficient and the bandwidth safety margin defined in the flow characteristics of historical industrial control data streams.

[0128] Specifically, if the overall performance deviation is positive (most indicators fail to meet the target), the adjustment direction is to increase bandwidth; if the overall deviation is negative (most indicators exceed the target), the adjustment direction is to compress bandwidth.

[0129] For example, if the baseline bandwidth requirement is 100Mbps, the bandwidth adjustment factor is 1.4, and the overall performance deviation is positive, then the adjusted bandwidth is 100Mbps × 1.4 = 140Mbps; if the bandwidth adjustment factor is 0.9, and the overall performance deviation is negative, then the adjusted bandwidth is 100Mbps × 0.9 = 90Mbps.

[0130] Furthermore, the bandwidth safety margin is a constraint threshold set to avoid over-adjustment of bandwidth. It is determined by the traffic characteristics of historical industrial control data flows. For example, if it is set to 30% of the baseline bandwidth, the adjusted bandwidth must not exceed 1.3 times or fall below 0.7 times the baseline bandwidth. If the baseline bandwidth is 100Mbps and the bandwidth adjustment factor is 1.4, the adjusted bandwidth can only be up to 130Mbps, not 140Mbps, due to the safety margin constraint. If the bandwidth adjustment factor is 0.6, the adjusted bandwidth can only be as low as 70Mbps, not 60Mbps, due to the safety margin constraint. By combining this multiplicative adjustment with the safety margin constraint, the targeted nature of the bandwidth adjustment is ensured, while network stability issues caused by over-adjustment are avoided, ultimately achieving an accurate match between the baseline bandwidth requirement and the transmission performance target.

[0131] After adjusting the baseline bandwidth requirement proportionally, the adjustment results are comprehensively verified against the bandwidth compatibility requirements of key production processes to calculate the updated bandwidth resource reservation value. For example, if the baseline bandwidth of 100Mbps is adjusted to 130Mbps by a bandwidth adjustment factor of 1.3, and meets the process's minimum limit of 80Mbps and maximum limit of 200Mbps, then this value is determined to be the final bandwidth resource reservation value.

[0132] Furthermore, the updated bandwidth resource reservation value is sent to the TSN switch via an encrypted configuration command. The command clearly specifies the transmission queue identifier and reservation value parameters to ensure accurate identification and execution by the switch. Simultaneously, the traffic shaping parameters corresponding to the transmission queue are adjusted to match the bandwidth resource reservation value, thereby ensuring stable and controlled traffic transmission rates.

[0133] Finally, by querying the switch configuration feedback information and monitoring the transmission performance indicators in real time, the parameter adjustment is confirmed to be effective until the transmission performance meets the preset target, thus completing the dynamic reconfiguration of bandwidth resources.

[0134] Furthermore, based on the time window replanning recommendations in the adjustment strategy, and using the time synchronization mechanism as a foundation, the available time window boundaries of the transmission queue are recalculated, a corresponding scheduling table is generated, and distributed to network devices to complete the adjustment of the time window boundaries. The time synchronization mechanism must ensure that all relevant devices in the 5G-TSN converged network have the same time base, for example, by achieving microsecond-level synchronization based on the IEEE 1588 PTP protocol, providing time support for time window planning.

[0135] Specifically, when calculating time window boundaries, it is necessary to consider the cyclical characteristics and transmission requirements of key production processes. For example, for cyclically scheduled processes with a command sending cycle of 10 milliseconds, the time window can be set to allocate a dedicated transmission period of 5 milliseconds every 10 milliseconds. Real-time control processes have higher timeliness requirements and can be planned with a 3-millisecond time window every 5 milliseconds. At the same time, it is necessary to avoid overlapping of different queue windows. For example, if a high-priority queue has a window of 0-3 milliseconds, a low-priority queue should start allocating a window from 3.5 milliseconds, reserving a 0.5-millisecond buffer.

[0136] In addition, the generated scheduling table needs to clearly define parameters such as queue identifier, window start time, and window duration, for example, "Queue 004: Start time 0ms, Duration 5ms, Repetition period 10ms". This table is then sent to the TSN switch and 5G base station via an encrypted protocol. Upon receiving the table, the device updates its local scheduling rules, strictly controls data transmission according to the new window boundaries, and simultaneously monitors transmission latency and jitter in real time to confirm the window adjustment is effective, ensuring the deterministic timing of data transmission.

[0137] Furthermore, after completing the full-process parameter reconfiguration of transmission queue priority update, bandwidth resource reservation value adjustment and time window boundary optimization, the reconfigured transmission performance indicators are continuously collected and used as the core feedback data for closed-loop optimization. The entire process from performance indicator monitoring to dynamic parameter reconfiguration is iteratively executed until the transmission of control command streams for key production processes meets the preset deterministic performance targets.

[0138] Among them, the transmission performance indicators mentioned above need to continue the previous monitoring dimensions, including the real-time and average values ​​of end-to-end latency, the fluctuation range of latency jitter, the cumulative percentage of packet loss rate, etc., to ensure the consistency and comparability of the data. For example, after reconfiguration, end-to-end latency data is collected every 1 millisecond, and 1000 sets are collected continuously to calculate the average latency and jitter amplitude.

[0139] Specifically, during each iteration, the feedback transmission performance metrics are compared one by one with the preset deterministic performance targets to determine whether all core metrics meet the requirements. For example, if the preset end-to-end latency target is 20 milliseconds, the latency jitter target is 2 milliseconds, and the packet loss rate target is 0.1%, and the average latency after reconfiguration is 18 milliseconds, the jitter is 1.5 milliseconds, and the packet loss rate is 0.08%, then the targets are considered met. If the average latency is 22 milliseconds and the jitter is 2.5 milliseconds, failing to meet the preset standards, then the feedback data needs to be re-input into the channel parameter dynamic optimization model. The model will then combine the effect of this reconfiguration to correct the priority allocation, bandwidth adjustment coefficient, or time window planning scheme in the adjustment strategy, generating a new round of parameter adjustment instructions.

[0140] Meanwhile, reasonable iteration cycles and termination conditions need to be set during iterative execution to avoid invalid loops. The iteration cycle can be set according to the response characteristics of the production process. For example, the iteration cycle for real-time control processes is 5 seconds to ensure rapid response to performance fluctuations; the iteration cycle for periodic scheduling processes is 30 seconds to balance optimization efficiency and system stability.

[0141] In addition, there are two types of termination conditions: First, if the core performance indicators consistently meet the preset targets for multiple consecutive iterations without significant fluctuations, such as the average latency being between 19 and 20 milliseconds for five consecutive monitoring cycles, the jitter not exceeding 2 milliseconds, and the packet loss rate remaining between 0.05% and 0.1%, then the iteration should be stopped. Second, if the number of iterations reaches the preset upper limit, and the performance indicators still do not meet the targets at this time, the iteration should be paused and the model parameters or network hardware problems should be investigated, optimized, and then the iteration process should be restarted.

[0142] Ultimately, through the aforementioned mechanism of continuous feedback and iterative optimization, the key parameters of the deterministic transmission channel can dynamically adapt to changes in production process requirements and fluctuations in network operation status. For example, when a production process switch causes a surge in traffic, the bandwidth reservation value can be quickly increased and the time window allocation optimized through iterative adjustments to ensure that the transmission performance remains stable within the preset range, providing continuous and reliable transmission assurance for the efficient and stable operation of critical production processes.

[0143] The embodiments of this application, through the above specific implementation methods, achieve the following technical effects: This application proposes a 5G-driven AI-adaptive production optimization method for digital factories. First, a 5G-TSN converged network infrastructure layer matching the target production area is constructed. 5G base stations and TSN switches are deployed with a time synchronization mechanism configured. Transmission queues with different priorities and time windows are pre-configured, and initial bandwidth resources and routing strategies are allocated to form an initial deterministic transmission channel resource pool. Simultaneously, historical industrial control data flows are analyzed, and multi-type non-IP industrial protocol command flow samples are parsed and labeled. A semantic understanding and encapsulation model for multiple industrial protocols is constructed and trained. Then, transmission queues are allocated from the initial resource pool for key production processes, and initial bandwidth and routing are configured to establish deterministic transmission channels. The semantic understanding and encapsulation model is used to parse control command flows and encapsulate them according to a preset format to generate standardized data units with timestamps, channel identifiers, and priority tags. Subsequently, the operational status data of the deterministic transmission channels and the end-to-end transmission performance indicators of the data units are monitored in real time and input into a pre-constructed dynamic optimization model for channel parameters. The model combines equipment characteristics and protocol configuration constraints to conduct collaborative analysis, quantifying the optimization cost and value, and outputting adjustment strategies including queue priority reallocation, bandwidth readjustment, and time window replanning. Finally, based on the adjustment strategy, the key parameters of the channel are dynamically reconfigured, the queue priority label is updated, the traffic shaping parameters and bandwidth reservation values ​​of the TSN switch are adjusted, the time window scheduling table is recalculated, and the performance indicators after reconfiguration are used as feedback to iteratively execute the entire process from monitoring to reconfiguration until the transmission of key production process control command streams meets the preset deterministic performance targets.

[0144] The method provided in this application, through the technical solution of "dual construction of network and model - channel establishment and data encapsulation - performance monitoring and strategy generation - parameter reconfiguration and iterative optimization", solves the problems of multi-protocol adaptation barriers, fixed and rigid channel parameters, imbalance between transmission performance and resource consumption, and lack of closed-loop optimization mechanism in traditional industrial transmission. It avoids transmission delay exceeding the standard, jitter aggravation or resource waste caused by dynamic changes in production process and network load fluctuations, and improves the determinism, real-time performance and stability of key production process control command transmission in digital factories.

[0145] Example 2, as shown in the appendix Figure 2 As shown, based on the inventive concept of the 5G-driven AI adaptive production optimization method for digital factories provided in Embodiment 1, this application also provides a 5G-driven AI adaptive production optimization system for digital factories, specifically including: The network infrastructure construction and model training module 01 is used to construct a 5G-TSN converged network base layer that matches the target production area, and to train a semantic understanding and encapsulation model for multiple industrial protocols by combining historical industrial control data streams. The channel establishment and protocol encapsulation module 02 is used to establish a deterministic transmission channel for key production processes in the 5G-TSN converged network base layer, and based on the semantic understanding and encapsulation model, encapsulate the control command stream of at least one non-IP industrial protocol involved in the key production process into a standardized data unit that can be transmitted through the deterministic transmission channel. The performance monitoring and strategy output module 03 is used to monitor the operating status data of the deterministic transmission channel and the end-to-end transmission performance indicators of the encapsulated data units in real time, input the operating status data and transmission performance indicators into the pre-built channel parameter dynamic optimization model, and output the adjustment strategy for the key parameters of the current channel. The parameter reconfiguration and iterative optimization module 04 is used to dynamically reconfigure the key parameters of the deterministic transmission channel according to the adjustment strategy, and use the reconfigured transmission performance index as feedback to iteratively execute the entire process from monitoring performance index to dynamic reconfiguration until the transmission of control command stream of the key production process meets the preset deterministic performance target.

[0146] In one embodiment, the network base construction and model training module 01 is further used for: Based on the physical distribution and communication relationships of industrial equipment in the target production area, 5G base stations and TSN switches are deployed to form a 5G-TSN converged network infrastructure layer covering the target production area. A time synchronization mechanism for the 5G-TSN converged network infrastructure layer is configured, and transmission queues with different priorities and time windows are divided and pre-configured according to the periodicity and timeliness constraints of the key production processes. The key production processes include periodically scheduled processes, event-driven processes, and real-time control processes. Based on the traffic characteristics of historical industrial control data streams, bandwidth resources and routing strategies are initially allocated to the transmission queues in the 5G-TSN converged network infrastructure layer to form an initial deterministic transmission channel resource pool.

[0147] Furthermore, the network infrastructure construction and model training module 01 also includes: Based on the historical industrial control data stream, instruction stream samples containing at least one non-IP industrial protocol are collected, and the protocol fields, control semantics, and instruction timing relationships in the instruction stream samples are parsed and labeled to obtain a set of labeled industrial protocol instruction samples. Based on machine learning, the initial network architecture of the semantic understanding and encapsulation model is constructed. The initial network architecture is trained under supervision using the set of industrial protocol instruction samples until the verification accuracy converges, thus completing the construction of the semantic understanding and encapsulation model.

[0148] In one embodiment, the channel establishment and protocol encapsulation module 02 is further used for: From the initial deterministic transmission channel resource pool, at least one transmission queue with corresponding priority and time window is allocated to the key production process. Based on the traffic characteristics of the historical industrial control data stream, initial bandwidth resources and initial routing strategies are configured for the transmission queue to establish a corresponding deterministic transmission channel. Using the semantic understanding and encapsulation model, the non-IP industrial protocol control command streams involved in the key production process are parsed and semantically understood in real time to extract the protocol command structure, field semantics, and control logic. Based on the transmission requirements of the deterministic transmission channel, the semantically understood control commands are adapted and encapsulated to generate standardized data units with timestamps, channel identifiers, and priority tags.

[0149] Furthermore, the channel establishment and protocol encapsulation module 02 also includes: The control commands are provided with a transmission time reference based on a time synchronization mechanism, and a timestamp is generated based on the transmission time reference. Priority tags are determined and labeled according to the priority and time window of the transmission queue allocated to the key production process, and the unique identifier of the deterministic transmission channel is used as the channel identifier. Based on the data unit length constraints determined by the flow characteristics of the historical industrial control data stream, the control commands are segmented or combined to form a data payload that meets transmission requirements. The timestamp, channel identifier, priority tag, and data payload are combined and encapsulated according to a preset format to generate a standardized data unit that can be transmitted through the deterministic transmission channel.

[0150] In one embodiment, the performance monitoring and strategy output module 03 is further used for: The system collects real-time operational status data of the deterministic transmission channel, including queue occupancy rate of the transmission queues allocated to the key production process, time deviation of the time synchronization mechanism, and real-time status of relevant links in the 5G-TSN converged network infrastructure layer. It also collects real-time end-to-end transmission performance indicators of the encapsulated data units, including end-to-end latency, latency jitter, and packet loss rate. The collected operational status data and transmission performance indicators are input into the pre-built dynamic optimization model for channel parameters. This model combines the equipment characteristics of the industrial equipment involved in the key production process with the protocol configuration uniformity constraints based on the semantic understanding and encapsulation model for collaborative analysis. In this collaborative analysis, the additional latency caused by protocol adaptation and encapsulation is quantified as the optimization cost, and the latency reduction brought about by optimized transmission channel configuration is quantified as the optimization value. Based on the optimization cost and optimization value, with minimizing overall latency and meeting the preset deterministic performance target as the optimization direction, an adjustment strategy for the current key parameters of the channel is output. This adjustment strategy includes at least one of the following: reallocation of transmission queue priorities, readjustment of initial bandwidth resources, or replanning of the time window.

[0151] In one embodiment, the parameter reconfiguration and iterative optimization module 04 is further used for: Based on the redistribution suggestion for transmission queue priority in the adjustment strategy, the priority label of the transmission queue bound to the deterministic transmission channel is updated by calling the network management interface of the 5G-TSN converged network base layer; based on the readjustment suggestion for initial bandwidth resources in the adjustment strategy, the bandwidth resource reservation value corresponding to the transmission queue is updated by adjusting the traffic shaping parameters of the transmission queue in the TSN switch; based on the replanning suggestion for time windows in the adjustment strategy, the available time window boundary of the transmission queue is adjusted by recalculating and issuing the scheduling table based on the time synchronization mechanism.

[0152] Furthermore, the parameter reconfiguration and iterative optimization module 04 also includes: Based on the traffic characteristics of the historical industrial control data stream, the baseline bandwidth requirements corresponding to the key production process are determined; the end-to-end transmission performance indicators of the encapsulated data units are acquired in real time; the end-to-end latency, latency jitter, and packet loss rate are compared with the preset deterministic performance target to calculate the performance deviation; combined with the optimization cost and optimization value obtained in the collaborative analysis, the latency trade-off state under the current network configuration is evaluated, and the baseline bandwidth requirements are proportionally adjusted based on the performance deviation to calculate the updated bandwidth resource reservation value; the updated bandwidth resource reservation value is sent to the TSN switch through configuration commands, and the traffic shaping parameters corresponding to the transmission queue are adjusted to complete the dynamic reconfiguration of bandwidth resources.

[0153] Furthermore, the parameter reconfiguration and iterative optimization module 04 also includes: The optimization value is compared with the optimization cost, and it is determined whether the current network configuration is in an effective optimization range based on whether the optimization value exceeds the optimization cost; the values ​​and signs of each deviation index in the performance deviation are obtained, and the deviation type and degree of deviation that dominate the performance deviation are identified; the bandwidth adjustment coefficient is determined based on the determination result of the effective optimization range and the degree of deviation of the dominant deviation type; based on the bandwidth adjustment coefficient, the baseline bandwidth requirement is multiplicatively adjusted, wherein the adjustment direction is determined by the overall sign of the performance deviation, and the adjustment magnitude is jointly constrained by the bandwidth adjustment coefficient and the bandwidth safety margin defined in the traffic characteristics of the historical industrial control data stream.

[0154] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.

[0155] The above description is only a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

[0156] This specification and accompanying drawings are merely illustrative examples of this application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from its scope. Therefore, if such modifications and variations fall within the scope of this application and its equivalents, this application intends to include such modifications and variations.

Claims

A 1.5G-driven AI-adaptive production optimization method for digital factories, characterized in that: The method includes: Construct a 5G-TSN converged network base layer that matches the target production area, and train a semantic understanding and encapsulation model for multiple industrial protocols by combining historical industrial control data streams. In the 5G-TSN converged network base layer, deterministic transmission channels are established for key production processes, and based on the semantic understanding and encapsulation model, control command streams of at least one non-IP industrial protocol involved in the key production processes are encapsulated into standardized data units that can be transmitted through the deterministic transmission channels. The system monitors the operational status data of the deterministic transmission channel and the end-to-end transmission performance indicators of the encapsulated data units in real time. It inputs the operational status data and transmission performance indicators into a pre-built dynamic optimization model for channel parameters and outputs adjustment strategies for the key parameters of the current channel. According to the adjustment strategy, the key parameters of the deterministic transmission channel are dynamically reconfigured, and the reconfigured transmission performance indicators are used as feedback to iteratively execute the entire process from monitoring performance indicators to dynamic reconfiguration until the transmission of control command streams for the key production process meets the preset deterministic performance target.

2. The 5G-driven AI adaptive production optimization method for digital factories according to claim 1, characterized in that, The construction steps of the 5G-TSN converged network base layer include: Based on the physical distribution and communication relationship of industrial equipment in the target production area, 5G base stations and TSN switches are deployed to form a 5G-TSN converged network infrastructure layer covering the target production area. Configure the time synchronization mechanism of the 5G-TSN converged network base layer, and divide and pre-configure transmission queues with different priorities and time windows according to the periodicity and timeliness constraints of the key production processes. The key production processes include periodic scheduling processes, event-driven processes, and real-time control processes. Based on the traffic characteristics of the historical industrial control data streams, bandwidth resources and routing strategies are initially allocated to the transmission queues in the 5G-TSN converged network base layer to form an initial deterministic transmission channel resource pool.

3. The 5G-driven AI adaptive production optimization method for digital factories according to claim 1, characterized in that, The steps for constructing the semantic understanding and encapsulation model include: Based on the historical industrial control data stream, instruction stream samples containing at least one non-IP industrial protocol are collected, and the protocol fields, control semantics, and instruction timing relationships in the instruction stream samples are parsed and annotated to obtain an annotated set of industrial protocol instruction samples. Based on machine learning, the initial network architecture of the semantic understanding and encapsulation model is constructed; The initial network architecture is trained under supervised supervision using the industrial protocol instruction sample set until the verification accuracy converges, thus completing the construction of the semantic understanding and encapsulation model.

4. The 5G-driven AI adaptive production optimization method for digital factories according to claim 1, characterized in that, Based on the semantic understanding and encapsulation model, the control command stream of at least one non-IP industrial protocol involved in the key production process is encapsulated into standardized data units that can be transmitted through the deterministic transmission channel, including: From the initial deterministic transmission channel resource pool, allocate at least one transmission queue with corresponding priority and time window for the key production process, and configure initial bandwidth resources and initial routing strategies for the transmission queue based on the traffic characteristics of the historical industrial control data stream, and establish a corresponding deterministic transmission channel. Using the semantic understanding and encapsulation model, the non-IP industrial protocol control command flow involved in the key production process is parsed and semantically understood in real time, and the protocol command structure, field semantics and control logic are extracted. Based on the transmission requirements of the deterministic transmission channel, the semantically understood control instructions are adapted and encapsulated to generate standardized data units with timestamps, channel identifiers, and priority tags.

5. The 5G-driven AI adaptive production optimization method for digital factories according to claim 4, characterized in that, Based on the transmission requirements of the deterministic transmission channel, the semantically understood control commands are adapted and encapsulated to generate standardized data units with timestamps, channel identifiers, and priority tags, including: The control command is provided with a transmission time reference according to the time synchronization mechanism, and the timestamp is generated based on the transmission time reference. Based on the priority and time window of the transmission queue allocated to the critical production process, the priority label is determined and marked, and the unique identifier of the deterministic transmission channel is used as the channel identifier. Based on the data unit length constraints determined by the flow characteristics of the historical industrial control data stream, the control commands are segmented or combined to form a data payload that meets the transmission requirements. The timestamp, channel identifier, priority tag, and data payload are combined and encapsulated according to a preset format to generate a standardized data unit that can be transmitted through a deterministic transmission channel.

6. The 5G-driven AI adaptive production optimization method for digital factories according to claim 1, characterized in that, The system monitors the operational status data of the deterministic transmission channel and the end-to-end transmission performance indicators of the encapsulated data units in real time. It then inputs this operational status data and transmission performance indicators into a pre-constructed dynamic optimization model for channel parameters, outputting adjustment strategies for key parameters of the current channel, including: Real-time acquisition of operational status data of the deterministic transmission channel, including queue occupancy rate of the transmission queue allocated to the key production process, time deviation of the time synchronization mechanism, and real-time status of relevant links in the 5G-TSN converged network infrastructure layer. Real-time acquisition of end-to-end transmission performance indicators of the encapsulated data units, including end-to-end latency, latency jitter, and packet loss rate; The collected operating status data and transmission performance indicators are input into the pre-constructed channel parameter dynamic optimization model, wherein the channel parameter dynamic optimization model combines the equipment characteristics of the industrial equipment involved in the key production process with the protocol configuration uniformity constraints based on the semantic understanding and encapsulation model to perform collaborative analysis; In the aforementioned collaborative analysis, the additional latency caused by protocol adaptation and encapsulation is quantified as the optimization cost, and the latency reduction brought about by optimized transmission channel configuration is quantified as the optimization value. Based on the optimization cost and optimization value, with the optimization direction of minimizing overall latency and meeting the preset deterministic performance target, an adjustment strategy for the current channel key parameters is output. The adjustment strategy includes at least one of the following: reallocation of transmission queue priority, readjustment of initial bandwidth resources, or replanning of time window.

7. The 5G-driven AI adaptive production optimization method for digital factories according to claim 1, characterized in that, According to the adjustment strategy, the key parameters of the deterministic transmission channel are dynamically reconfigured, including: According to the reallocation suggestion for transmission queue priority in the adjustment strategy, the priority label of the transmission queue bound to the deterministic transmission channel is updated by calling the network management interface of the 5G-TSN converged network base layer. Based on the readjustment recommendations for initial bandwidth resources in the adjustment strategy, the bandwidth resource reservation value corresponding to the transmission queue is updated by adjusting the traffic shaping parameters of the transmission queue in the TSN switch. Based on the time window replanning recommendations in the adjustment strategy, the available time window boundaries of the transmission queue are adjusted by recalculating and issuing the scheduling table based on the time synchronization mechanism.

8. The 5G-driven AI adaptive production optimization method for digital factories according to claim 7, characterized in that, Based on the readjustment recommendations for initial bandwidth resources in the adjustment strategy, the bandwidth resource reservation value corresponding to the transmission queue is updated by adjusting the traffic shaping parameters of the transmission queue in the TSN switch, including: Based on the flow characteristics of the historical industrial control data stream, the baseline bandwidth requirements corresponding to the key production process are determined. Obtain the end-to-end transmission performance indicators of the encapsulated data units collected in real time; The end-to-end latency, latency jitter, and packet loss rate are compared with the preset deterministic performance target to calculate the performance deviation. Based on the optimization costs and values ​​derived from the collaborative analysis, the latency trade-offs under the current network configuration are evaluated, and the baseline bandwidth requirement is proportionally adjusted based on the performance deviation to calculate the updated bandwidth resource reservation value. The updated bandwidth resource reservation value is sent to the TSN switch via configuration commands, and the traffic shaping parameters corresponding to the transmission queue are adjusted to complete the dynamic reconfiguration of bandwidth resources.

9. The 5G-driven AI adaptive production optimization method for digital factories according to claim 8, characterized in that, Based on the optimization costs and values ​​derived from the collaborative analysis, the latency trade-offs under the current network configuration are evaluated, and the baseline bandwidth requirement is proportionally adjusted based on the performance deviations, including: The optimization value is compared with the optimization cost, and it is determined whether the current network configuration is in an effective optimization range based on whether the optimization value exceeds the optimization cost. Obtain the values ​​and signs of each deviation index in the performance deviation, and identify the deviation type and degree of deviation that dominate the performance deviation; Based on the determination result of the effective optimization interval and the degree of deviation from the dominant deviation type, the bandwidth adjustment coefficient is determined; Based on the bandwidth adjustment coefficient, the baseline bandwidth requirement is multiplicatively adjusted, wherein the adjustment direction is determined by the overall sign of the performance deviation, and the adjustment magnitude is jointly constrained by the bandwidth adjustment coefficient and the bandwidth safety margin defined in the flow characteristics of the historical industrial control data stream. The 10.5G-driven AI-adaptive production optimization system for digital factories is characterized by: The system is used to execute the 5G-driven AI adaptive production optimization method for digital factories according to any one of claims 1-9, the system comprising: The network infrastructure construction and model training module is used to build a 5G-TSN converged network base layer that matches the target production area, and to train a semantic understanding and encapsulation model for multiple industrial protocols by combining historical industrial control data streams. The channel establishment and protocol encapsulation module is used to establish a deterministic transmission channel for key production processes in the 5G-TSN converged network base layer, and based on the semantic understanding and encapsulation model, encapsulate the control command stream of at least one non-IP industrial protocol involved in the key production processes into a standardized data unit that can be transmitted through the deterministic transmission channel. The performance monitoring and strategy output module is used to monitor the operating status data of the deterministic transmission channel and the end-to-end transmission performance indicators of the encapsulated data units in real time, input the operating status data and transmission performance indicators into the pre-built channel parameter dynamic optimization model, and output adjustment strategies for the key parameters of the current channel. The parameter reconfiguration and iterative optimization module is used to dynamically reconfigure the key parameters of the deterministic transmission channel according to the adjustment strategy, and use the reconfigured transmission performance index as feedback to iteratively execute the entire process from monitoring performance index to dynamic reconfiguration until the transmission of control command stream of the key production process meets the preset deterministic performance target.

Images