Digital twin industrial switch remote state monitoring management system

Abstract

The application relates to the technical field of industrial automation and network communication, and particularly discloses a digital twin industrial switch remote state monitoring management system, which comprises an edge agent module, a digital twin modeling and synchronization engine module and a visual interaction module. The edge agent module collects and evaluates the state of a physical switch, dynamically decides the data sampling frequency, compression algorithm and redundant sending strategy in combination with the network condition fed back by the cloud. The cloud module updates a high-fidelity digital twin model and feeds back the network quality, and simultaneously performs predictive analysis. Through adaptive synchronization and intelligent decision-making, the application realizes reliable and real-time synchronization of key state data under network fluctuation, and improves the real-time performance and early warning capability of state monitoring.

Description

Technical Field

[0001] This invention belongs to the field of industrial automation and network communication technology, and specifically relates to a remote status monitoring and management system for digital twin industrial switches. Background Technology

[0002] Industrial networks serve as the core support for critical infrastructure in modern intelligent manufacturing, energy and power, and traffic control, and their stable and reliable operation is of paramount importance. Industrial switches, as key nodes in the network, are responsible for data forwarding and traffic control, and their operational status directly affects the performance and security of the entire industrial network. Real-time and accurate status monitoring and management of industrial switches has become a fundamental requirement for ensuring the stable operation of industrial systems.

[0003] Digital twin technology offers a new paradigm for the virtual mapping and status analysis of industrial equipment. By constructing a digital twin model of a physical switch, remote visual monitoring and historical data analysis of its operational status can be achieved. The basic principle of this technology is to continuously collect operational data from the physical switch through sensors and communication interfaces and synchronize it to its digital twin model, thereby realizing the mirroring and simulation of the physical entity in virtual space.

[0004] Existing technologies typically employ periodic data collection and transmission to report status data such as port traffic, CPU load, and temperature of physical switches to their digital twin models. However, in complex industrial network environments, network bandwidth fluctuations and link congestion are commonplace, posing challenges to this fixed sampling rate synchronization mechanism. Increased data synchronization latency causes the status information in the digital twin model to lag significantly behind the actual status of the physical switches.

[0005] This latency prevents the system from providing real-time warnings and responses to critical anomalies such as sudden overloads and buffer overflows at physical switch ports. Furthermore, when network bandwidth is limited, high-frequency data streams at fixed frequencies are highly susceptible to network congestion, leading to increased packet loss rates and compromised data integrity upon which the digital twin model is based, resulting in misjudgments of state. Therefore, ensuring the real-time and reliable synchronization of state data between the digital twin model and physical switches under fluctuating industrial network conditions has become a pressing technical challenge. Summary of the Invention

[0006] The purpose of this invention is to provide a remote status monitoring and management system for digital twin industrial switches, in order to solve the technical contradiction that under the conditions of industrial network bandwidth fluctuation and link congestion, the real-time performance and reliability of status data synchronization between physical switches and their digital twin models are poor, resulting in the status lag of the digital twin model and the failure of abnormal early warning.

[0007] To achieve the above objectives, this invention provides a remote status monitoring and management system for digital twin industrial switches. The system includes an edge agent module deployed on the physical industrial switch side, a digital twin modeling and synchronization engine module deployed on a cloud or local server, and a visualization and interaction module deployed on a management terminal.

[0008] The edge proxy module is used to reside inside the physical industrial switch to perform data acquisition, local status assessment, and adaptive data synchronization strategy decision-making.

[0009] This module further includes a data acquisition submodule, a status assessment submodule, and a synchronization strategy decision submodule.

[0010] The data acquisition submodule collects raw status data from multiple hardware monitoring points and software interfaces of the physical switch at millisecond intervals.

[0011] These monitoring points include at least the real-time inbound and outbound byte rates, packet rates, error frame counts, packet loss counts, CPU load percentage and memory usage of the switching chip, temperature sensor readings of key internal components of the device, and specific alarm events in the system log for each network port.

[0012] The state assessment submodule is used to process and extract features from the collected raw state data in real time, and calculate the comprehensive state deviation index based on the preset multi-level state judgment rules.

[0013] The process begins by standardizing the raw data to eliminate dimensional differences.

[0014] Subsequently, the standardized data is compared with the preset static baseline threshold to identify preliminary abnormal dimensions.

[0015] Furthermore, the state assessment submodule introduces a dynamic baseline calculation unit, which calculates the dynamic normal range of various state parameters based on data within a historical time window using a moving average algorithm.

[0016] The state assessment submodule compares the current data with the static baseline and the dynamic baseline simultaneously, and assigns a deviation weighting coefficient to each state parameter based on the comparison results.

[0017] Finally, through a weighted summation algorithm, the deviations of all state parameters are aggregated into a comprehensive state deviation index between 0 and 100, where 0 represents completely normal and 100 represents severe abnormality.

[0018] The synchronization strategy decision submodule is the core of this invention for achieving adaptive synchronization. The inputs to this submodule are the comprehensive state deviation index output by the state assessment submodule, and the network status feedback parameters periodically issued by the digital twin modeling and synchronization engine module.

[0019] The network condition feedback parameters characterize the average network round-trip latency and packet loss rate experienced by data during the previous synchronization cycle when it is uploaded to the cloud.

[0020] The synchronization strategy decision-making submodule internally maintains a two-dimensional decision matrix. One dimension of this matrix is ​​a comprehensive state deviation index, which is divided into four discrete levels: normal, attention, warning, and emergency.

[0021] Another dimension is the network status feedback parameter, which is divided into three discrete levels: smooth, mild congestion, and severe congestion.

[0022] Each cell of the decision matrix contains a predefined set of complete data synchronization strategy parameter combinations.

[0023] This combination specifically includes: data sampling frequency, data compression algorithm selection identifier, and the number of redundant transmissions of critical status data.

[0024] For example, when the overall status deviation index is at the "normal" level and the network status is "smooth", the strategy parameters are low-frequency sampling, lossless compression enabled, and single transmission.

[0025] When the overall status deviation index enters the "early warning" level and the network status is "mild congestion", the strategy parameters are adjusted to medium frequency sampling, lossy compression is enabled but key features are retained, and two redundant transmissions are used to ensure that key alarm information arrives reliably.

[0026] The synchronization strategy decision submodule queries the decision matrix based on the real-time input status and network parameters, dynamically generates and executes the optimal data synchronization command for the current moment, and controls the content, frequency and method of the data packets sent by the edge agent module to the cloud.

[0027] The digital twin modeling and synchronization engine module is used to receive and process status data uploaded by the edge agent module in the cloud, drive the operation and update of the high-fidelity digital twin model, and calculate network status feedback parameters.

[0028] This module further includes a model-driven submodule, a synchronous quality assessment submodule, and a predictive analytics submodule.

[0029] The model-driven submodule maintains a high-fidelity digital twin model that corresponds one-to-one with the physical switch.

[0030] This model not only includes the three-dimensional geometric appearance of the physical switch, but also integrates its internal logical forwarding model, traffic processing model, and thermodynamic model.

[0031] Upon receiving the status data packet uploaded by the edge agent, the model-driven submodule first decompresses the data packet according to the compression algorithm identifier carried in the packet header to restore the complete data.

[0032] Subsequently, these data are used to update the corresponding parameters in the digital twin model in real time, such as updating the virtual port traffic count, adjusting the load simulation value of the virtual switching chip, and recalculating the virtual temperature field distribution inside the device.

[0033] The updated model status is provided to the visualization and interaction module in real time via the application programming interface.

[0034] The synchronization quality assessment submodule is responsible for monitoring the performance of the data synchronization link. This submodule maintains a session context for each active edge proxy connection, recording the sequence number, timestamp, and data integrity verification results of data packets received in a recent period.

[0035] By analyzing the jitter in the arrival interval of data packets, the continuity of sequence numbers, and the failure of verification, this submodule calculates network condition feedback parameters that characterize the quality of the current network path. These parameters include quantified round-trip delay estimates and packet loss rate estimates.

[0036] This parameter is encapsulated in the downlink control command of the next cycle and sent to the corresponding edge agent module for use by its synchronization strategy decision submodule.

[0037] The predictive analytics submodule runs a time series prediction algorithm based on the historical state data sequence accumulated by the digital twin model.

[0038] The algorithm uses a long short-term memory network model, taking the state data of the previous N time steps as input, to predict the development trend of key state indicators, such as core port traffic and central processing unit load, within the next M time steps.

[0039] When the predicted value exceeds the set safety threshold, the predictive analysis submodule generates a predictive alarm event. This event not only includes the alarm content, but also the associated state parameters that led to the prediction result and their historical change curves, which are then pushed to the visualization interaction module.

[0040] The visualization and interaction module provides system administrators with a unified monitoring and management interface. This module obtains real-time updated 3D digital twin model status, real-time alarm lists, historical trend charts, and predictive alarm information from the digital twin modeling and synchronization engine module.

[0041] Administrators can use this interface to drill down into the digital twin of a single switch from a global network topology perspective, and view its real-time internal and external operating status, including the device temperature distribution displayed in the form of a heat map, and the data flow direction and traffic volume between ports displayed in the form of a dynamic flow graph.

[0042] All alarms and predictions are highlighted in relation to specific components in the 3D model.

[0043] As one embodiment of the present invention, the specific workflow of the dynamic baseline calculation unit in the state assessment submodule is as follows: For each monitored state parameter, the dynamic baseline calculation unit maintains a first-in-first-out data buffer of fixed length L, where L is 1024.

[0044] Whenever a new parameter value is acquired, it is inserted at the end of the buffer, and the oldest data is removed. The dynamic baseline calculation unit periodically performs statistical analysis on the data in the buffer, calculating its mean and standard deviation.

[0045] The upper limit of the dynamic normal range is set as the mean plus K times the standard deviation, and the lower limit is set as the mean minus K times the standard deviation, where K is a preset sensitivity coefficient with a typical value of 2.

[0046] If the current parameter value exceeds this dynamic range, it is considered a dynamic anomaly and is assigned a higher deviation weight coefficient when calculating the comprehensive state deviation index.

[0047] As one embodiment of the present invention, the specific generation logic of the strategy parameters of the two-dimensional decision matrix in the synchronization strategy decision submodule follows the following principles: When the overall status deviation index increases, the data sampling frequency and redundant transmission times are gradually increased to ensure that abnormal statuses are captured and reported intensively and reliably.

[0048] When the network congestion level indicated by the network condition feedback parameters worsens, a compression algorithm with a higher compression ratio is activated. It may also filter out some non-critical steady-state monitoring data without affecting the anomaly detection, thereby reducing the bandwidth consumption of the synchronous data stream and preventing the reported data itself from exacerbating network congestion.

[0049] This design allows for a dynamic and optimal balance between the "quality" and "quantity" of data synchronization, taking into account the urgency of the physical object's state and the congestion of the transmission channel.

[0050] As one embodiment of the present invention, the model-driven submodule in the digital twin modeling and synchronization engine module updates the thermodynamic model of the high-fidelity digital twin model as follows: the temperature sensor data received by the model-driven submodule is mapped to the corresponding geometric region of the heating element inside the three-dimensional model.

[0051] The thermodynamic model is based on the principle of finite element analysis, which discretizes the internal space of the equipment into small mesh units.

[0052] Based on the real-time power consumption data of each heat-generating element, a heat source is applied to the corresponding grid cell.

[0053] The model solves the transient heat conduction equation based on physical parameters such as the thermal conductivity and convective heat transfer coefficient of the material, calculates the temperature value of each grid cell in the next time step, and generates a dynamically updated temperature field distribution map within the entire 3D model, which is then presented in the visualization interface through color gradients.

[0054] In one embodiment of the present invention, the system further includes a strategy optimization closed-loop module. This module periodically collects historical operation logs, which record the strategy parameters adopted by the synchronization strategy decision submodule for each instance, the overall state deviation index at that time, the network status feedback parameters, and the actual effect of the synchronization data packet successfully arriving at the cloud and updating the model.

[0055] The strategy optimization closed-loop module uses a reinforcement learning algorithm to jointly optimize the combination of strategy parameters in each cell of the two-dimensional decision matrix within the synchronization strategy decision submodule. This enables the system to adapt to the constantly changing specific industrial network environment and continuously improve synchronization efficiency and reliability.

[0056] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. This invention fundamentally changes the traditional fixed-frequency data reporting mode by introducing an adaptive synchronization mechanism that combines edge intelligence and cloud collaboration. The system assesses its own urgency in real time at the physical switch side and dynamically adjusts data sampling, compression, and transmission strategies based on network quality information fed back from the cloud. This allows the system to operate in a low-frequency, low-bandwidth mode when the device is stable and the network is unobstructed; once an anomaly is detected or a risk is predicted, the priority and reliability of data reporting are immediately increased, ensuring that critical anomaly information can penetrate network congestion and reach the digital twin model in real time. This effectively solves the problem of lag in the digital twin model's state under network fluctuations, improving the real-time performance of state monitoring and the ability to respond to sudden anomalies.

[0057] 2. This invention constructs a complete technical closed loop from data acquisition, status assessment, strategy decision-making to model updates. The status assessment submodule employs a multi-dimensional judgment method combining static and dynamic baselines, enabling more sensitive and accurate identification of potential abnormal trends, rather than simply responding to threshold alarms. The digital twin model is not merely a passive display of status; its integrated predictive analysis submodule can proactively warn of future risks based on historical data, shifting the operation and maintenance model from post-event remediation to pre-event prevention. The visualization interaction module provides a high-fidelity, three-dimensional, data-correlated visualization interface, improving administrators' cognitive efficiency and decision-making accuracy regarding the internal status of complex equipment and network behavior.

[0058] 3. This invention, through the design of a strategy optimization closed-loop module, endows the system with the ability to self-evolve and continuously optimize. The system can utilize historical operational data and automatically optimize its core synchronous decision matrix using reinforcement learning algorithms, allowing its strategy parameter combinations to continuously adapt to specific industrial network environments and equipment operating characteristics. This design enables the system to not only provide performance superior to fixed strategies in the initial deployment phase, but also to become increasingly efficient and intelligent over time through continuous learning. This achieves long-term evolution of the monitoring and management system's adaptive capabilities, ensuring the long-term applicability and reliability of the technology in complex and ever-changing industrial scenarios. Attached Figure Description

[0059] Figure 1 This is a schematic diagram of the overall technical solution architecture of the digital twin industrial switch remote status monitoring and management system proposed in this invention; Figure 2 This is a schematic diagram of the core principle framework of the adaptive data synchronization strategy decision based on state and network conditions in this invention; Figure 3 This is a flowchart illustrating the logical process framework for the edge agent module to perform local state assessment and decision-making in this invention. Figure 4 This is a schematic diagram of the multi-level interaction and data flow of the digital twin modeling and synchronization engine module in this invention for model updating and network evaluation; Figure 5 This is a schematic diagram of the core principle framework for the strategy optimization closed-loop module in this invention to achieve synchronous strategy self-evolution. Detailed Implementation

[0060] Please refer to the attached document. Figures 1 to 5 This embodiment details a specific implementation plan for a remote status monitoring and management system for digital twin industrial switches. The system aims to address the core technical challenge of poor real-time performance and low reliability of status data synchronization between physical industrial switches and their corresponding cloud-based digital twin models under conditions of dynamic bandwidth fluctuations and intermittent link congestion in industrial network environments.

[0061] The overall system architecture consists of three main parts working together: an edge agent module deployed inside the physical industrial switch hardware, a digital twin modeling and synchronization engine module deployed in a cloud data center or a local high-performance server, and a visualization and interaction module deployed on the system administrator's work terminal.

[0062] These three modules interact bidirectionally with data and control commands through the Industrial Internet or dedicated network channels, forming a complete closed loop from physical layer data perception to cloud-based intelligent processing and human-machine interaction decision-making.

[0063] First, the edge agent module deployed on the side of the physical industrial switch is described in detail.

[0064] This module resides directly in the operating system kernel or user space of the physical switch in software form. Its core responsibilities are to perform local data acquisition with millisecond-level precision, conduct real-time status assessment and health measurement, and dynamically decide and execute the optimal data synchronization strategy based on the assessment results and network quality information fed back from the cloud.

[0065] The edge proxy module is logically forced to be split into three technically tightly coupled and sequentially executed sub-components: a data acquisition sub-module, a status assessment sub-module, and a synchronization strategy decision sub-module. This split ensures modularity of functions and clarity of processing flow.

[0066] The data acquisition submodule is the starting point for the system to perceive the physical world.

[0067] This submodule collects raw status data synchronously from multiple key hardware monitoring points and software log interfaces of the device by directly calling the underlying driver interface and system management information database interface of the physical switch, with a configurable millisecond-level sampling period.

[0068] These monitoring points constitute a comprehensive monitoring network for the operating status of the switch, specifically including at least: for each physical network port on the device, real-time collection of its inbound byte rate, outbound byte rate, inbound data packet rate, outbound data packet rate, received error frame count, sent error frame count, and packet loss count due to buffer fullness. For the core processing unit of the switch, namely the switching chip or central processing unit, collect its load percentage and system memory usage. For critical components inside the equipment used to ensure stable operation, such as the surface of the switching chip, the vicinity of the power module, and the fan exhaust vent, the embedded temperature sensor readings are collected. These values ​​are usually in degrees Celsius. In addition, the data acquisition submodule continuously monitors the system logs and captures and extracts specific alarm event entries, such as link oscillation logs, authentication failure logs, or configuration change logs, by matching predefined regular expression patterns.

[0069] All collected raw data is appended with a high-precision timestamp and data source identifier, and temporarily stored in a thread-safe circular buffer, awaiting further processing by the state evaluation submodule.

[0070] The status assessment submodule is responsible for real-time processing, feature extraction, and comprehensive health assessment of the raw status data provided by the data acquisition submodule.

[0071] The core output of this submodule is a quantitative comprehensive state deviation index, with a value range of 0 to 100, where 0 represents that the equipment is in a completely normal state and 100 represents that the equipment is in a seriously abnormal or faulty state.

[0072] To achieve this goal, the processing flow within the state assessment submodule is further refined into four consecutive technical stages: data standardization preprocessing, static baseline threshold comparison, dynamic baseline range calculation, and comprehensive index weighted aggregation.

[0073] In the first stage, the data standardization preprocessing unit receives raw data from the circular buffer.

[0074] Different state parameters have different units and numerical ranges. For example, the port rate is measured in megabits per second and the value may be as high as thousands, while the CPU load is measured in percentages and the value ranges from 0 to 100, and the temperature reading is measured in degrees Celsius and the value range is smaller.

[0075] To eliminate the impact of dimensional differences on subsequent comprehensive evaluation, the preprocessing unit applies either the min-max normalization algorithm or the Z-score standardization algorithm based on historical statistics to each parameter, mapping the raw data to a uniform, dimensionless numerical range of 0 to 1. The standardized data are denoted as the standardized parameter values.

[0076] In the second stage, the static baseline threshold comparison unit begins operation. The system has a pre-set static baseline threshold table, which defines the upper and lower limits of the normal operating range of various status parameters under factory or health calibration conditions.

[0077] For example, the static baseline upper limit for inbound traffic of a gigabit Ethernet port may be preset to 800 megabits per second, the static baseline upper limit for CPU load may be preset to 85%, and the static baseline upper limit for chip junction temperature may be preset to 95 degrees Celsius.

[0078] The static baseline threshold comparison unit compares each standardized parameter value at the current time with its corresponding static baseline threshold range.

[0079] If a parameter value exceeds the static baseline threshold range, the parameter is immediately marked as a "static anomaly," and its deviation is recorded, i.e., the absolute distance of the parameter value from the threshold boundary.

[0080] In the third phase, the dynamic baseline computing units work in parallel.

[0081] The purpose of this unit is to capture the normal change patterns caused by natural fluctuations in business load during equipment operation, thereby more sensitively identifying potential abnormal trends that deviate from historical normal patterns and avoiding false alarms during normal business peaks.

[0082] For each monitored status parameter, the dynamic baseline calculation unit maintains a fixed-length first-in-first-out data buffer of 1024 in memory.

[0083] Whenever a new parameter value is acquired, it is inserted at the end of the buffer, and the oldest historical data point at the head of the buffer is automatically removed, thus always keeping the buffer able to hold data from the most recent 1024 sampling periods.

[0084] The dynamic baseline calculation unit periodically, for example every 60 seconds, performs statistical analysis on all 1024 data points in the current buffer and calculates their arithmetic mean. with standard deviation .

[0085] Subsequently, the dynamic baseline calculation unit calculates the sensitivity coefficient based on the preset sensitivity coefficient. With a typical value set to 2, the upper and lower limits of the dynamic normal range of this parameter within the current time window are calculated. The upper limit of the dynamic normal range is... The lower limit of the dynamic normal range is .

[0086] If the parameter value at the current moment exceeds this dynamic range, it will be marked as "dynamic anomaly".

[0087] The dynamic baseline calculation unit outputs the calculated upper and lower limits of the dynamic range, as well as the deviation of the current parameter value from the dynamic range, to the next stage.

[0088] In the fourth stage, the comprehensive index weighted aggregation unit receives the results from static baseline comparison and dynamic baseline calculation.

[0089] This unit maintains a weight coefficient configuration table, which assigns a deviation weight coefficient to each state parameter in both static and dynamic anomaly cases.

[0090] Typically, dynamic anomalies are given a slightly higher weight than static anomalies because dynamic anomalies are more likely to represent a newly emerging potential failure trend that deviates from historical patterns.

[0091] For each parameter, select the corresponding weight coefficient based on whether it triggers a static or dynamic exception.

[0092] If both are triggered simultaneously, the higher of the two values ​​is used. Subsequently, the aggregation unit calculates the weighted sum of deviations for all parameters.

[0093] Specifically, for the first The standardized deviation of the parameter is denoted as . The corresponding weighting coefficient is denoted as Overall deviation index The calculation follows the formula: ; The total number of state parameters participating in the evaluation. The function ensures that the final indicator is less than the upper limit of 100. Through the refined processing of the above four stages, the status assessment submodule can output a comprehensive quantitative health indicator that reflects both immediate exceedances and trend deviations.

[0094] The synchronization strategy decision submodule is the core hub for the edge agent module to achieve intelligent adaptive synchronization. This submodule has two inputs: First, the comprehensive state deviation index is output in real time by the state assessment submodule; Second, network status feedback parameters are periodically issued by the digital twin modeling and synchronization engine module.

[0095] Network condition feedback parameters are structured data objects that encapsulate the quality assessment results of the network path from this edge agent to the cloud server during the previous data synchronization cycle. They primarily include two quantitative metrics: Average network round-trip latency estimate in milliseconds; and uplink data packet loss rate estimate in percentage.

[0096] The core of the synchronization strategy decision submodule is a predefined and online-updable two-dimensional decision matrix. Please refer to the appendix. Figure 2 One dimension of this matrix is ​​discretized into four levels based on the numerical range of the comprehensive state deviation index S: Normal level correspondence Pay attention to the corresponding level The warning level corresponds to Emergency level correspondence .

[0097] The other dimension of the matrix is ​​determined based on a combination of round-trip delay and packet loss rate in the network condition feedback parameters, and is discretized into 3 levels: The smooth flow level corresponds to a round-trip time of less than 50 milliseconds and a packet loss rate of less than 0.1%. The mild congestion level corresponds to a round-trip time of between 50 and 200 milliseconds or a packet loss rate of between 0.1% and 5%. The severe congestion level corresponds to a round-trip time of more than 200 milliseconds or a packet loss rate of more than 5%.

[0098] Each cell of this two-dimensional decision matrix predefines a complete set of data synchronization strategy parameters. This set of parameters specifically controls three key synchronization behaviors: Data sampling frequency, data compression algorithm selection identifier, and redundant transmission count of critical status data.

[0099] The data sampling frequency determines the cycle in which the edge agent obtains data from the data acquisition submodule and triggers a state assessment and synchronization decision. For example, low frequency corresponds to 1 second, medium frequency corresponds to 200 milliseconds, and high frequency corresponds to 50 milliseconds.

[0100] The data compression algorithm selection flag is used to specify the compression algorithm used before encapsulating status data into network packets. For example, flag 0 means no compression, flag 1 means lossless compression algorithm such as LZ4 is used, and flag 2 means lossy compression algorithm is used, but core feature fields such as comprehensive status deviation index, critical port traffic, central processor load, and temperature over-limit value will be retained first before compression, while some slow-changing or non-critical monitoring data such as error count details of all ports will be filtered out.

[0101] Redundancy transmission count defines the number of times the same critical data packet is repeatedly transmitted at the network layer to cope with network packet loss, such as single transmission, double redundancy transmission, or triple redundancy transmission.

[0102] The real-time decision-making process of the synchronization strategy decision-making submodule is as follows: Whenever the state assessment submodule generates a new comprehensive state deviation index or receives new network status feedback parameters from the cloud, the decision-making submodule initiates a decision query.

[0103] It first determines the status level based on the value of the comprehensive status deviation index; for example, 65 belongs to the warning level.

[0104] At the same time, it analyzes the latest network status feedback parameters and determines the network status level based on round-trip latency and packet loss rate. For example, a round-trip latency of 150 milliseconds and a packet loss rate of 2% indicate a mild congestion level.

[0105] Subsequently, the decision submodule queries the two-dimensional decision matrix with status level as the row and network condition level as the column, and locates the corresponding cell.

[0106] Assuming the query result is a cell representing the intersection of the warning level and the mild congestion level, the predefined strategy parameters within that cell are: The sampling frequency is 200 milliseconds, the compression algorithm identifier is 2 (lossy compression retains key features), and the number of redundant transmissions is 2.

[0107] The decision submodule immediately generates control instructions containing these parameters and sends them to the data acquisition submodule and the data encapsulation and sending thread.

[0108] The data acquisition submodule then adjusts its sampling period to 200 milliseconds. After new data is acquired and evaluated, the encapsulation thread selects the corresponding lossy compressor based on identifier 2 to process the data, and then copies the generated data packet into two copies, which are then sent to the cloud sequentially via independent network sockets or at very short intervals.

[0109] Through this mechanism, the system achieves dynamic adaptation of the synchronization strategy: when the device is in a stable state and the network is good, it adopts a low-frequency, lossless, single-transmission energy-saving low-bandwidth mode. If the equipment deteriorates or the network quality declines, the system will automatically improve the real-time performance of data reporting and enhance the robustness of data transmission to ensure that critical anomaly information can be synchronized to the cloud in a priority and reliable manner.

[0110] Next, we will describe in detail the digital twin modeling and synchronization engine module deployed in the cloud.

[0111] This module runs on a cloud server cluster with high computing performance and reliable storage. It is responsible for receiving and processing status data uploaded from all edge agent modules across the network, driving the real-time operation and accurate updates of the high-fidelity digital twin model, and evaluating network conditions to form a feedback loop. Logically, this module is forcibly divided into three core sub-components: a model-driven sub-module, a synchronous quality assessment sub-module, and a predictive analytics sub-module.

[0112] The model-driven submodule is the core engine of the digital twin world. It maintains a one-to-one high-fidelity digital twin model instance for each physical industrial switch on the network.

[0113] This model is far more than just a display of three-dimensional geometry; it is a simulation system that deeply integrates multiphysics and logical behavior.

[0114] Specifically, the model integrates at least the following sub-models: A logical forwarding model is used to simulate the MAC address table learning, packet forwarding decision-making, and virtual LAN isolation logic of a switch. Traffic processing model used to simulate traffic statistics, queue scheduling and congestion control behavior of various network ports; A thermodynamic model is used to simulate the heat transfer and distribution processes generated by the power consumption of electronic components inside the device. Each twin model instance is associated with a physical switch through a unique device identifier.

[0115] The processing flow begins when the model-driven submodule receives a data packet from the edge agent via the network interface.

[0116] First, the submodule parses the packet header and extracts the compression algorithm identifier. Based on this identifier, it calls the corresponding decompression algorithm to restore the complete structured state data.

[0117] These data are categorized and mapped to the corresponding parameters of the twin model.

[0118] For example, the port inbound byte rate is used to update the virtual port traffic count and drive the traffic processing model to recalculate the port's load rate; The CPU load percentage is used to adjust the simulated load value of the virtual switching chip, which may affect the simulated processing latency of the logical forwarding model. The readings from each temperature sensor are input into the thermodynamic model. Updating the thermodynamic model is a computationally intensive process.

[0119] Please refer to the attached document. Figure 4 The model-driven submodule maps the received temperature data onto the corresponding geometric regions of the heating elements inside the 3D model. These heating elements include switching chips, memory chips, power conversion modules, etc.

[0120] The thermodynamic model is based on the principle of finite element analysis, which pre-discretes the entire internal space and shell of the switch into hundreds of thousands or even millions of tiny tetrahedral or hexahedral mesh elements. Each heating element is associated with a set of mesh elements.

[0121] The model applies an equivalent heat source intensity to the associated grid cells based on the element's real-time power consumption data.

[0122] Subsequently, the model solves the three-dimensional unsteady heat conduction equation based on predefined physical parameters such as the thermal conductivity, convective heat transfer coefficient, and radiation coefficient of each component material, and calculates the temperature value of each grid cell in the next simulation time step.

[0123] This process ultimately generates a dynamically updated, continuous temperature field distribution map within the entire 3D model.

[0124] The updated model status, including updated port traffic, chip load, temperature field data, and 3D model attitude information, is pushed in real time to all visualization and interactive module instances that subscribe to the device status via a high-performance WebSocket or gRPC application interface.

[0125] The synchronization quality assessment submodule is responsible for monitoring and evaluating the communication quality of each data synchronization link from the edge agent to the cloud.

[0126] This submodule maintains an independent session context data structure for each active edge proxy connection.

[0127] Within this context, detailed information about the 256 most recently received data packets is recorded, including packet sequence numbers, the precise timestamps of when the packets were received in the cloud, and data integrity verification results. The synchronization quality assessment submodule periodically analyzes these records.

[0128] Network jitter can be assessed by calculating the variance of the time intervals between consecutive data packets. Packet loss events can be identified by checking the continuity of packet sequence numbers.

[0129] By counting the number of failed integrity checks, the bit error rate during transmission can be assessed.

[0130] Based on these analyses, this submodule uses algorithms such as moving average and Kalman filtering to calculate network status feedback parameters that characterize the current network path quality.

[0131] This parameter includes two core estimates: the quantified average round-trip time estimate and the uplink packet loss rate estimate.

[0132] After the calculation is completed, the parameter is encapsulated into a specific field of the next downlink heartbeat packet or control command data packet and sent to the corresponding edge agent module as an important input to its synchronization strategy decision submodule, thereby completing the network quality feedback loop from the cloud to the edge.

[0133] The predictive analytics submodule is dedicated to enabling operations and maintenance to shift from reactive response to proactive prevention.

[0134] This submodule uses historical state data time series accumulated from digital twin models as a basis to run advanced time series prediction algorithms.

[0135] Specifically, for key status indicators of each switch, such as traffic on the core uplink ports, CPU load, and core chip temperature, the predictive analytics submodule maintains a dedicated long short-term memory network model.

[0136] The model uses historical state data from the previous 128 consecutive time steps as input features. After calculation by a complex internal gated loop unit, it outputs a sequence of predicted values ​​for the indicator over the next 32 time steps. The predictive analysis submodule continuously compares the predicted values ​​with a preset safety threshold. When more than a certain proportion, such as 70%, of the data points in the predicted sequence exceeds the safety threshold, the module immediately generates a predictive alarm event.

[0137] The event includes not only the alarm level, alarm content, and predicted occurrence time, but also a list of associated state parameters that led to the prediction and their historical change curves over a period of time.

[0138] This rich contextual information, along with predictive alerts, is pushed to the visualization and interaction module in real time, providing administrators with in-depth decision support.

[0139] Finally, the visual interaction module deployed on the administrator's work terminal is described in detail.

[0140] This module is typically presented as a web application or desktop client, providing system administrators with a unified monitoring and management interface for the entire industrial switch network and individual device details.

[0141] The visualization and interaction module obtains and presents multi-dimensional information in real time by subscribing to the application programming interface provided by the digital twin modeling and synchronization engine module.

[0142] The interface typically uses a multi-view layout, including a global network topology view, a device list view, a detailed view of a single device's 3D digital twin, an alarm and event panel, and a historical data analysis view.

[0143] In the global network topology view, all industrial switches are presented in simplified icon form of their digital twin model. The lines between the icons represent physical or logical network connections, and the line width and color dynamically reflect the real-time traffic volume and link health status.

[0144] Administrators can click on any switch, and the interface will drill down to a detailed 3D digital twin view of that switch.

[0145] In this view, administrators can rotate, zoom, and slice the high-fidelity 3D model of the switch 360 degrees.

[0146] The virtual temperature field distribution inside the equipment is superimposed on the model surface and internal cross-section in the form of a heat map, with the color gradually changing from blue to red, intuitively showing the areas where heat accumulates.

[0147] The data flow between network ports is displayed in the form of dynamic, particle flow or strip flow graphs of varying thicknesses, making the direction of the flow and bandwidth utilization immediately apparent.

[0148] When a real-time or predictive alarm occurs, the alarm and event panel will immediately pop up a notification, and the specific components related to the alarm in the 3D model will automatically highlight and flash. For example, the chip area that generates an over-temperature alarm will flash red, and the port that generates a traffic overload alarm will flash orange.

[0149] When an administrator clicks on an alarm item, the interface can display historical trend charts, correlation parameter curves, and predictive analysis reports related to that alarm, realizing deep visualization of the correlation between data, models, and alarms, which greatly improves the efficiency of fault location and root cause analysis.

[0150] This embodiment further illustrates the key enhancement features of the system: Strategy optimization closed-loop module. Please refer to the appendix. Figure 5 This module, as an optional advanced component of the digital twin modeling and synchronization engine module, empowers the system with the ability to learn on its own and continuously optimize its synchronization strategy.

[0151] The strategy optimization closed-loop module periodically, for example once a day, collects historical operational data from the system log database.

[0152] Each log entry records a complete synchronization event, and its fields include at least: The data includes timestamps, device identifiers, the combination of synchronization strategy parameters used by the edge agent at the time, the comprehensive state deviation index at the time of decision-making, the network status feedback parameters used at the time of decision-making, whether the synchronization data packet successfully arrived at the cloud, the end-to-end latency from data collection to model update, and whether the data captured the real abnormal state that was later confirmed.

[0153] The policy optimization closed-loop module uses these historical logs as training datasets to build a reinforcement learning environment.

[0154] In this environment, the agent is the synchronous policy decision submodule. Its action space is all possible combinations of policy parameters in the two-dimensional decision matrix, and its state space is the Cartesian product of the comprehensive state deviation index level and the network condition level.

[0155] The reward function is designed with two core objectives in mind: First, minimize the overall end-to-end latency of state synchronization; second, maximize the capture rate of real abnormal states.

[0156] The policy optimization closed-loop module uses reinforcement learning algorithms such as deep Q-networks or policy gradients to learn and evaluate historical decision sequences offline.

[0157] Through continuous iteration, the algorithm learns which combination of policy parameters yields higher long-term cumulative rewards under specific state and network conditions. The learned new combination of policy parameters is used to update the global policy library maintained in the cloud, and is incrementally synchronized to the synchronous policy decision submodule of the corresponding edge agent module through a secure downlink channel, replacing the old policy in the corresponding cell of its two-dimensional decision matrix.

[0158] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0159] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A remote status monitoring and management system for digital twin industrial switches, characterized in that, include: The edge agent module, deployed on the side of the physical industrial switch, is used to reside inside the physical industrial switch to perform data acquisition, local status assessment, and adaptive data synchronization strategy decision-making. The digital twin modeling and synchronization engine module, deployed in the cloud or on a local server, is used to receive and process status data uploaded by the edge agent module, drive the operation and update of the high-fidelity digital twin model, and calculate network status feedback parameters. The visualization and interaction module deployed on the management terminal is used to obtain real-time updated 3D digital twin model status, real-time alarm list, historical trend charts and predictive alarm information from the digital twin modeling and synchronization engine module, and provides a unified monitoring and management interface.

2. The remote status monitoring and management system for a digital twin industrial switch according to claim 1, characterized in that, The edge agent module includes a data acquisition submodule, a status assessment submodule, and a synchronization strategy decision submodule. The data acquisition submodule is used to collect raw status data from multiple hardware monitoring points and software interfaces of the physical industrial switch at millisecond intervals. The state assessment submodule is used to process and extract features from the collected raw state data in real time, and calculate a comprehensive state deviation index between 0 and 100 based on preset multi-level state judgment rules. The synchronization strategy decision submodule takes as input the comprehensive state deviation index output by the state assessment submodule and the network status feedback parameters periodically issued by the digital twin modeling and synchronization engine module. The submodule maintains a two-dimensional decision matrix. One dimension of the matrix is ​​the discrete level of the comprehensive state deviation index, and the other dimension is the discrete level of the network status feedback parameters. Each cell of the decision matrix predefines a set of data synchronization strategy parameter combinations. The synchronization strategy decision submodule queries the decision matrix based on the real-time input status and network parameters, dynamically generates and executes the data synchronization command for the current moment.

3. The remote status monitoring and management system for a digital twin industrial switch according to claim 2, characterized in that, The digital twin modeling and synchronization engine module includes a model-driven submodule, a synchronization quality assessment submodule, and a predictive analysis submodule; The model-driven submodule is used to maintain a high-fidelity digital twin model that corresponds one-to-one with the physical switch. After receiving the status data packet uploaded by the edge agent, it decompresses the data packet according to the compression algorithm identifier carried in the header and uses the restored data to update the corresponding parameters in the digital twin model in real time. The synchronization quality assessment submodule is used to maintain the session context for each active edge proxy connection. By analyzing the received data packet sequence number, timestamp, and data integrity verification result, it calculates the network condition feedback parameters, which include the quantized round-trip delay estimate and packet loss rate estimate, and encapsulates these parameters in the downlink control command of the next cycle and sends them to the corresponding edge proxy module. The predictive analysis submodule is used to run a time series prediction algorithm based on the historical state data sequence accumulated by the digital twin model to predict the development trend of key state indicators in future time steps, and generate a predictive alarm event when the predicted value is greater than the set safety threshold.

4. The remote status monitoring and management system for a digital twin industrial switch according to claim 3, characterized in that, The process of real-time processing and feature extraction of the raw state data by the state assessment submodule includes: Standardize all raw data to eliminate dimensional differences; The standardized data is compared with a preset static baseline threshold to identify preliminary abnormal dimensions. A dynamic baseline calculation unit is introduced, which calculates the dynamic normal range of various state parameters based on data within a historical time window using a moving average algorithm. The current data is compared with the static baseline and the dynamic baseline simultaneously, and a deviation weighting coefficient is assigned to each state parameter based on the comparison results. By using a weighted summation algorithm, the deviations of all state parameters are aggregated into the comprehensive state deviation index.

5. The remote status monitoring and management system for a digital twin industrial switch according to claim 4, characterized in that, The workflow of the dynamic baseline calculation unit is as follows: For each monitored status parameter, maintain a fixed-length first-in-first-out data buffer; Whenever a new parameter value is collected, it is inserted at the end of the buffer, and the oldest data is removed; Periodically perform statistical analysis on the data in the buffer, and calculate its mean and standard deviation; The mean plus K times the standard deviation is set as the upper limit of the dynamic normal range, and the mean minus K times the standard deviation is set as the lower limit of the dynamic normal range, where K is the preset sensitivity coefficient. If the current parameter value exceeds this dynamic range, it is considered a dynamic anomaly and is given a higher deviation weight coefficient when calculating the comprehensive state deviation index.

6. The remote status monitoring and management system for a digital twin industrial switch according to claim 5, characterized in that, In the synchronization strategy decision submodule, the discrete level division of the two-dimensional decision matrix is ​​as follows: The overall status deviation index is divided into four discrete levels: normal, attention, warning, and emergency. The network status feedback parameters are divided into three discrete levels: smooth, mild congestion, and severe congestion. The specific combination of data synchronization strategy parameters includes: data sampling frequency, data compression algorithm selection identifier, and redundant transmission count of key status data.

7. The remote status monitoring and management system for a digital twin industrial switch according to claim 6, characterized in that, The generation logic of strategy parameters in the synchronization strategy decision submodule follows the following principles: When the overall state deviation index increases, the data sampling frequency and the number of redundant transmissions should be increased. When the network congestion level, as represented by the network condition feedback parameters, worsens, a compression algorithm with a higher compression ratio is activated, and some non-critical steady-state monitoring data is filtered out without affecting the anomaly detection.

8. The remote status monitoring and management system for a digital twin industrial switch according to claim 7, characterized in that, The high-fidelity digital twin model maintained by the model-driven submodule integrates a logical forwarding model, a traffic processing model, and a thermodynamic model. The update process of the thermodynamic model is as follows: the received temperature sensor data is mapped to the corresponding geometric region of the heating element inside the three-dimensional model; Based on the principle of finite element analysis, the internal space of the device is discretized into grid cells; according to the real-time power consumption data of each heating element, a heat source is applied to the corresponding grid cell; based on the physical parameters of the material's thermal conductivity coefficient and convective heat transfer coefficient, the transient heat conduction equation is solved, and the temperature value of each grid cell in the next time step is calculated, thereby generating a dynamically updated temperature field distribution map.

9. A remote status monitoring and management system for a digital twin industrial switch according to claim 8, characterized in that, The time series prediction algorithm run by the predictive analysis submodule uses a long short-term memory network model. This long short-term memory network model uses the state data of the previous N time steps as input to predict the development trend of key state indicators in the next M time steps. The predictive alarm event includes alarm content, as well as the associated state parameters that lead to the prediction result and their historical change curves.

10. A remote status monitoring and management system for a digital twin industrial switch according to claim 9, characterized in that, The system also includes a strategy optimization closed-loop module; The strategy optimization closed-loop module is used to periodically collect historical operation logs. The logs record the strategy parameters adopted by the synchronization strategy decision submodule each time, the comprehensive state deviation index at that time, the network status feedback parameters, and the actual effect of the synchronization data packet successfully arriving at the cloud and updating the model. The strategy optimization closed-loop module uses a reinforcement learning algorithm to iteratively optimize the combination of strategy parameters in each cell of the two-dimensional decision matrix within the synchronization strategy decision submodule, with the joint optimization objective of minimizing the overall state synchronization delay and maximizing the abnormal state capture rate.

Images