Link self-healing topology control method of redundant communication network

Abstract

The invention relates to the technical field of communication network topology control, and provides a link self-healing topology control method for a redundant communication network, which realizes dynamic coupling evaluation of performance and environment by fusing the instantaneous deterioration degree and historical trend of performance data with the risk probability of environmental data, and improves the reliability of the communication network. The vulnerability index can truly reflect the current fault risk of the link, and a basis is provided for topology reconstruction triggering; after a high-fragility link is shielded, an alternative topological structure is explored through a graph theory algorithm, and it is ensured that the alternative topological structure adapts to service requirements; during service migration, service interruption caused by direct switching is avoided through parallel transmission and staged migration, and the requirement for continuity of high-real-time service is particularly met; by monitoring operation feedback data of a target topological path, vulnerability evaluation parameters and weighted evaluation parameters are dynamically updated, so that subsequent vulnerability index output and target topological structure selection can adapt to the actual operation state and service demand change of a link, and control precision attenuation caused by static parameters is avoided.

Description

Technical Field

[0001] This application relates to the field of communication network topology control technology, and in particular to a link self-healing topology control method for redundant communication networks. Background Technology

[0002] With increasing demands for communication reliability in fields such as industrial control and the Internet of Things (IoT), redundant communication networks have become a key supporting technology. By deploying multiple links and backup nodes, service continuity can be guaranteed in the event of link failures. Currently, topology control technologies for redundant communication networks mainly revolve around link quality assessment and topology control. Many of these methods assess link quality from a single dimension, such as collecting link performance data or environmental data, lacking dynamic adaptation. When link quality is poor, conventional methods use Dijkstra's algorithm to generate alternative topologies, but this is not integrated with risk control measures such as link shielding and path migration.

[0003] Existing technologies often assess link quality based solely on performance or environmental data, failing to adapt to the actual anti-interference capabilities of the link and the dynamic changes in services. This leads to a disconnect between link quality and the actual risk of link failure. When exploring alternative topologies, most only meet basic connectivity requirements without setting constraints on the interaction characteristics of node services, resulting in topologies that may not be suitable for service needs. During service migration, direct switching or indiscriminate parallel transmission can easily lead to interruptions of high-priority services, failing to meet real-time requirements. Furthermore, link quality assessment metrics are mostly statically set and not updated based on feedback data from actual topology operation, resulting in resource waste or insufficient performance in subsequently selected topologies.

[0004] To address the shortcomings of the existing technologies, the technical problem solved by this application is how to comprehensively evaluate link quality using multi-source data from redundant communication networks in complex scenarios in order to achieve self-healing topology control. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this application provides a link self-healing topology control method for redundant communication networks. The method includes: acquiring performance data and environmental data of the redundant communication network, fusing the instantaneous degradation degree and historical trend of the performance data with the risk probability of the environmental data, and dynamically calculating and outputting the vulnerability index of each link in the redundant communication network.

[0006] When the vulnerability index of any link exceeds the index threshold, topology reconfiguration is triggered. Based on the current topology, highly vulnerable links are shielded, and alternative topologies that meet connectivity requirements are explored in the redundant communication network using graph theory algorithms. The global robustness, service assurance, and reconfiguration cost of each alternative topology are evaluated in a weighted manner, and the target topology is selected based on the evaluation results.

[0007] The target topology path corresponding to the target topology structure is transmitted in parallel with the current topology path. During the parallel transmission process, the data stream of the current topology path is migrated to the target topology path in stages according to the service priority. After the migration, the link resources of the current topology path are released, and the operation feedback data of the target topology path is monitored to update the output vulnerability index and weighted evaluation parameters.

[0008] As an optional implementation, the output vulnerability index includes:

[0009] The system acquires performance and environmental data for each link in a redundant communication network, extracts features from the performance and environmental data, and obtains instantaneous degradation, historical trends, and risk probabilities.

[0010] By combining business load and normal operation benchmarks, the deviation between instantaneous degradation and historical trends is corrected to obtain performance correction characteristics that reflect link performance degradation. Based on the sensitivity of the link to environmental risk scenarios, the risk probability is converted into an environmental correction factor that affects performance degradation.

[0011] Based on historical fault data, establish the correlation strength coefficient between environmental risk scenarios and performance degradation, and adjust the weight ratio of the correlation strength coefficient according to the service priority of the current redundant communication network to form a fusion weight of performance correction characteristics and environmental correction factors.

[0012] The performance correction characteristics and environmental correction factors are nonlinearly fused according to the fusion weights to dynamically calculate and output the vulnerability index of each link in the redundant communication network.

[0013] As an optional implementation, the features of extracting performance data and environmental data include:

[0014] The instantaneous degradation of performance data is extracted by a sliding time window, and the historical trend of performance data is extracted based on a long short-term memory network.

[0015] The node locations and movement trajectories in the environmental data are converted into spatial topological coordinates, and the probability of the link being affected by environmental interference is calculated by combining the interference source information.

[0016] As an optional implementation, exploring the alternative topologies includes:

[0017] Based on the remaining links after shielding highly vulnerable links, and combined with the backup links and node resources of redundant communication networks, a topology resource pool is formed, labeled with link bandwidth, anti-interference capability and transmission delay.

[0018] Node groups are divided according to the service interaction characteristics of nodes in redundant communication networks, joint constraints are set for links between different node groups, and joint constraints are determined based on the labeling information of the topology resource pool.

[0019] Based on joint constraints, a graph theory algorithm with link redundancy constraints is used to explore paths in the topology resource pool, generate a basic topology that satisfies connectivity, and supplement redundant paths with adaptive annotation information to expand the topology and generate a set of candidate topologies.

[0020] Calculate the betweenness centrality of each link in each candidate topology, remove redundant links whose betweenness centrality is less than the betweenness threshold and do not satisfy the joint constraints, and obtain the candidate topology structure.

[0021] As an optional implementation, the shielding of highly vulnerable links includes:

[0022] When the vulnerability index of any link exceeds the index threshold, topology reconstruction is triggered, and all marked highly vulnerable links are identified based on the current topology.

[0023] Associate the connection role and service priority of highly vulnerable links in the current topology of redundant communication networks, and determine the support weight of highly vulnerable links for services based on service priorities;

[0024] The impact coefficient of highly vulnerable links on the connectivity of the current topology is determined based on the connection role, and the shielding impact level of highly vulnerable links is divided based on the support weight and the impact coefficient.

[0025] Assess the topological and service impact of shielding highly vulnerable links at each shielding impact level, shield highly vulnerable links sequentially, and determine whether to suspend the shielding operation based on the connectivity of the current topology during the shielding process.

[0026] As an optional implementation, the selection of the target topology includes:

[0027] Based on the annotation information of the topology resource pool, the backup links and node resources of the redundant communication network, and the service priority, the evaluation indicators including global robustness, service assurance and reconstruction cost are quantified.

[0028] The weights of the evaluation indicators are adjusted based on the proportion of redundant links and the proportion of spare resources in the alternative topology, forming an indicator weight matrix.

[0029] Simulate the random removal of non-highly vulnerable links from alternative topologies, calculate the decrease in connectivity retention rate and service assurance of alternative topologies after removal, and determine whether to add a risk resistance index.

[0030] The basic evaluation value of the candidate topology is obtained by weighting the index weight matrix, and the comprehensive evaluation value is obtained by superimposing the risk resistance index. The target topology is then selected based on the comprehensive evaluation value.

[0031] As an optional implementation, the updated vulnerability index includes:

[0032] After releasing the link resources of the current topology path, monitor the operation feedback data of the target topology path. The operation feedback data includes path operation parameters and service migration quality.

[0033] The anti-interference performance of the target topology path's path operation parameters is correlated and compared with the sensitivity of the link to environmental risk scenarios in order to adjust the environmental correction factor.

[0034] Degradation trends are extracted based on the time-series changes of path operation parameters and similarity matching is performed with historical fault data. The correlation strength coefficient between environmental risk scenarios and performance degradation is adjusted according to the matching results.

[0035] Based on the service migration quality of different services during parallel transmission, the fusion weights of performance correction features and environmental correction factors are adjusted accordingly.

[0036] Based on the adjusted environmental correction factor, correlation strength coefficient, and fusion weight, nonlinear fusion is re-executed to dynamically calculate and update the output vulnerability index.

[0037] As an optional implementation, the parallel transmission includes:

[0038] Establish parallel transmission channels between the target topology path and the current topology path, allocate transmission resources based on service priority, and initiate data flow migration of the current topology path in stages according to service priority;

[0039] Real-time monitoring and comparison of the path operation parameters of the two paths, and adjustment of the migration ratio of the data flow based on the degree of degradation of the target topology path relative to the current topology path;

[0040] Once the data flow of high-priority services has been fully migrated to the target topology path and the data flow of low-priority services has met the preset migration ratio, the link resources of the current topology path will be released in batches according to the connection role of the link.

[0041] As an optional implementation, updating the parameters of the weighted evaluation includes:

[0042] The link load fluctuation and anti-interference performance of the path operation parameters are correlated with global robustness; the remaining link reuse rate of the path operation parameters is correlated with reconstruction costs; and the quality of service migration is correlated with service assurance.

[0043] The evaluation metrics of the target topology during the selection phase are compared with the corresponding path operation parameters and service migration quality during the parallel transmission phase to obtain the deviation coefficient of the evaluation metrics.

[0044] The adjustment priority of the indicator weights is determined based on the magnitude of the deviation coefficient, and the indicator weights are adjusted in combination with the deviation coefficient to regenerate the indicator weight matrix for target topology selection.

[0045] Compared with existing technologies, the beneficial effects of this application are as follows: By integrating the instantaneous degradation degree of performance data and the risk probability of historical trends and environmental data, dynamic coupling assessment of performance and environmental factors is achieved, enabling the vulnerability index to truly reflect the current failure risk of the link and providing an accurate basis for triggering topology reconfiguration; after shielding highly vulnerable links, alternative topologies are explored through graph theory algorithms to ensure that alternative topologies adapt to business needs; during business migration, parallel transmission and phased migration are used to avoid service interruptions caused by direct switching, especially meeting the continuity requirements of high real-time services; by monitoring the operational feedback data of the target topology path, the vulnerability assessment parameters and weighted assessment parameters are dynamically updated, so that the subsequent vulnerability index output and target topology selection can adapt to the actual operating status of the link and changes in business needs, avoiding the control accuracy decay caused by static parameters. Attached Figure Description

[0046] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein:

[0047] Figure 1 This is a flowchart of a link self-healing topology control method for redundant communication networks provided in an embodiment of this application.

[0048] Figure 2 A logical flowchart illustrating the exploration of alternative topologies for the link self-healing topology control method for redundant communication networks provided in this application embodiment;

[0049] Figure 3 This is a logic flowchart of parallel transmission for the link self-healing topology control method of the redundant communication network provided in the embodiments of this application. Detailed Implementation

[0050] To make the objectives, technical solutions, and advantages of the embodiments of this application more apparent and understandable, 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.

[0051] like Figure 1 The diagram shown illustrates a method flowchart for a link self-healing topology control method for redundant communication networks, as provided in this application embodiment. The method includes:

[0052] This embodiment uses an industrial control redundant communication network as the application scenario. This scenario has stringent requirements for the stability and real-time performance of the link transmission. Fluctuations in link performance and environmental interference can easily cause abnormal transmission of production control commands.

[0053] S1. Acquire performance and environmental data of redundant communication networks, integrate the instantaneous degradation and historical trends of performance data with the risk probability of environmental data, dynamically calculate and output the vulnerability index of each link in the redundant communication network.

[0054] Specifically, the output vulnerability index includes:

[0055] The system acquires performance and environmental data for each link in a redundant communication network, extracts features from the performance and environmental data, and obtains instantaneous degradation, historical trends, and risk probabilities.

[0056] By combining business load and normal operation benchmarks, the deviation between instantaneous degradation and historical trends is corrected to obtain performance correction characteristics that reflect link performance degradation. Based on the sensitivity of the link to environmental risk scenarios, the risk probability is converted into an environmental correction factor that affects performance degradation.

[0057] Based on historical fault data, establish the correlation strength coefficient between environmental risk scenarios and performance degradation, and adjust the weight ratio of the correlation strength coefficient according to the service priority of the current redundant communication network to form a fusion weight of performance correction characteristics and environmental correction factors.

[0058] The performance correction characteristics and environmental correction factors are nonlinearly fused according to the fusion weights to dynamically calculate and output the vulnerability index of each link in the redundant communication network.

[0059] The vulnerability index needs to comprehensively reflect the overall state of link performance degradation and environmental interference. However, the raw data is disorganized, and direct use in calculations can distort the vulnerability index. Furthermore, the characteristics of different links need to be accurately distinguished; otherwise, feature-link mismatch will occur. Through a distributed acquisition and centralized aggregation model, performance and environmental data are acquired in real time by nodes at both ends of each link. Performance data includes transmission latency, packet loss rate, and link load, while environmental data includes node location, movement trajectory, and interference source information. After simple preprocessing by the nodes, the data from each link is centrally aggregated, categorized by link identifier, and the instantaneous degradation degree, historical trend, and risk probability of each link are extracted. The features are then bound to the corresponding link identifiers to form the features of each link. Distributed acquisition relies on nodes at both ends of the link to ensure real-time and comprehensive data acquisition, avoiding incomplete data coverage caused by single-point acquisition. Centralized aggregation and classification improve data quality, and link identifier binding ensures the correspondence between features and links, effectively avoiding feature mismatch risks and laying the foundation for subsequent accurate calculations.

[0060] The extracted performance characteristics are affected by fluctuations in service load. For example, during peak industrial production periods, the increased number of control commands carried by the link leads to increased load and consequently increased latency. This is not a performance degradation of the link itself, and using it directly will misjudge the link status. At the same time, the original risk probability only reflects the possibility of interference and does not quantify its actual impact on performance degradation, so it cannot be directly integrated with performance characteristics for calculation. When correcting performance deviations, a correlation between service load and normal operating benchmark is established in advance. For example, light load, medium load, and heavy load correspond to different normal ranges of transmission latency and packet loss rate. After determining the service load carried by the current link, the normal operating benchmark for the corresponding service load is called. The instantaneous degradation degree and historical trend are compared with the normal operating benchmark to eliminate performance fluctuations caused by load fluctuations. For example, if the latency increase under heavy load is within the range of the normal operating benchmark, it is not included in the link degradation itself, thus obtaining performance correction characteristics.

[0061] When converting the environmental correction factor, the sensitivity level is determined by combining the inherent characteristics of the physical medium of the link with historical fault data statistics. Wireless radio frequency links are inherently sensitive to electromagnetic interference and temperature and humidity fluctuations, wired Ethernet links are inherently sensitive to vibration and temperature fluctuations, and fiber optic links are inherently insensitive to various environmental risks. By analyzing historical fault data, the performance degradation of different links under different environmental risk scenarios is statistically analyzed to establish a correspondence between link type and the sensitivity level of environmental risk scenario. Specifically, if the failure rate of the same type of link in a certain environmental risk scenario exceeds 50%, it is judged as high sensitivity level, with a sensitivity level of 0.9; if the failure rate is between 20% and 50%, it is judged as medium sensitivity level, with a sensitivity level of 0.6; and if the failure rate is less than 20%, it is judged as low sensitivity level, with a sensitivity level of 0.3.

[0062] For example, wireless links experience a 65% failure rate under electromagnetic interference scenarios, classifying them as highly sensitive; ordinary wired links experience a 30% failure rate under temperature interference scenarios, classifying them as moderately sensitive; and fiber optic links experience a failure rate of less than 10% under any environmental risk scenario, classifying them as low sensitive. Risk probabilities are numerically categorized as high risk (≥0.7), medium risk (0.3-0.7), and low risk (≤0.3). Combining risk probability with sensitivity level maps to the environmental correction factor level. If a link is sensitive to a certain type of environmental risk scenario and has a high risk probability, then... The environment is converted into a high-value environmental correction factor to indicate that the environment has a significant impact on performance degradation, and vice versa. High sensitivity level and low risk, low sensitivity level and high risk are converted into a medium-value environmental correction factor. Performance deviation correction effectively eliminates the interference of business load fluctuations, so that performance characteristics more accurately reflect the degradation state of the link itself and avoid the problem of misjudging business load fluctuations as link degradation. The environmental correction factor conversion transforms the risk probability into a quantitative impact on performance degradation, realizes the dimensional unification of environmental factors and performance factors, and provides feasibility for subsequent fusion calculation of the two.

[0063] The correlation between environmental risk scenarios and performance degradation is not fixed. For example, if a certain type of electromagnetic interference has historically caused wireless link performance degradation multiple times, the correlation between the two is strong. Conventional methods use fixed coefficients, which cannot adapt to historical failure patterns in different scenarios. Different services have different priorities regarding performance and environment. For example, high-level services such as control command transmission are more concerned with performance stability, while ordinary data transmission services have higher requirements for environmental adaptability. When performing fusion calculations, the weights need to be adjusted according to service priorities to adapt to service needs. Based on historical failure data, a correlation strength coefficient between environmental risk scenarios and performance degradation is established. This involves screening out failure cases where performance degradation is caused by environmental interference, statistically analyzing the frequency and duration of link performance degradation under different environmental risk scenarios, and determining scenarios with high frequency and long duration as having a high correlation strength with performance degradation, and setting a higher correlation strength coefficient accordingly. Conversely, a lower correlation strength coefficient is set for scenarios with low frequency and long duration.

[0064] The weighting of the correlation strength coefficient is adjusted based on the service priorities of the current redundant communication network. Service priorities are predefined, with control command transmission classified as high-level, production data transmission as medium-level, and log data transmission as low-level. Initial rules for service priorities and weighting are established. For example, the fusion weight of performance correction features for high-level services (0.7) is higher than that of the environmental correction factor (0.3), while the fusion weight of the environmental correction factor for low-level services can be appropriately increased. Specifically, the fusion weight of performance correction features for low-level services is 0.3, and the fusion weight of the environmental correction factor is 0.7. Similarly, the fusion weights of performance correction features and the environmental correction factor for medium-level services are 0.5 each. The weighting of the correlation strength coefficient is adjusted according to these initial rules to form the fusion weight of performance correction features and the environmental correction factor. This establishes the correlation strength coefficient based on historical fault data, making the correlation between environment and performance more consistent with actual network operation patterns, avoiding discrepancies between theoretical and practical correlations. Adjusting the fusion weights in conjunction with service priorities makes the vulnerability index calculation more adaptable to the core needs of different services, improving the service adaptability of the vulnerability index.

[0065] Performance correction features and environmental correction factors have a complex coupling relationship. For example, strong electromagnetic interference can exacerbate link performance degradation, while links with high performance redundancy are more resistant to environmental interference. Linear fusion cannot reflect this coupling relationship, which can easily lead to distortion of the vulnerability index. At the same time, the link status changes in real time with service load and environmental interference, and the static index cannot reflect the current vulnerability status of the link in a timely manner. By using a fully connected neural network as a nonlinear fusion model, performance correction features and environmental correction factors are used as the input layer of the model, and the fusion weights are used as the connection weights from the input layer to the hidden layer. The hidden layer uses the ReLU activation function to handle the nonlinear coupling relationship between the two, such as simulating the dynamic process of environmental interference exacerbating performance degradation. The output layer outputs the vulnerability index of the link. The model is pre-trained by the correspondence between historical features and actual vulnerability status to ensure fusion accuracy. The actual vulnerability status includes whether the link has failed and the severity of the failure.

[0066] When link performance or environmental data changes, i.e., when interference sources appear or service loads switch, nodes upload new data in real time, triggering a dynamic recalculation of the entire process, including deviation correction, environmental correction factor conversion, fusion weight adjustment, and nonlinear fusion. The vulnerability index is updated in real time, and the vulnerability index of each link in the redundant communication network is stored and marked in real time. The fully connected neural network effectively captures the nonlinear coupling relationship between performance and environment through activation functions. Compared with linear fusion, it can more accurately reflect the comprehensive vulnerability status of the link. Dynamic calculation ensures that the vulnerability index can follow the changes in link status in real time, avoiding the problem that static indices cannot reflect real-time vulnerability, and providing an accurate basis for timely triggering of subsequent topology reconstruction.

[0067] Furthermore, the features extracted from performance data and environmental data include:

[0068] The instantaneous degradation of performance data is extracted by a sliding time window, and the historical trend of performance data is extracted based on a long short-term memory network.

[0069] The node locations and movement trajectories in the environmental data are converted into spatial topological coordinates, and the probability of the link being affected by environmental interference is calculated by combining the interference source information.

[0070] Conventional link performance evaluation only uses performance parameters at a single point in time, making it difficult to avoid the impact of occasional data fluctuations. Furthermore, the simple periodic statistical extraction of historical patterns fails to capture the long-term cumulative trend of performance degradation, resulting in only partial performance characteristics that cannot support a comprehensive vulnerability assessment. This approach selects link transmission latency, packet loss rate, and link load as performance data, relying on real-time acquisition of performance data from both ends of the link, with the sampling frequency consistent with the transmission cycle of industrial control services. A sliding time window is used to extract the instantaneous degradation degree of the performance data, where the window sliding step size is synchronized with the sampling frequency. By comparing the performance data within the window with a preset normal link performance range, the number of consecutive occurrences of degraded data within the window and the trend direction of deviation from the normal range are statistically analyzed to represent the instantaneous degradation degree.

[0071] Based on the Long Short-Term Memory (LSTM) network, historical trends in performance data are extracted. A sliding window of instantaneous degradation over a continuous one-hour period is used as the time-series input. After receiving the time-series data, the input layer of the LTM network filters historical features through gating units in two hidden layers, including the periodic occurrence of a certain type of degradation and the cumulative magnitude of degradation. Finally, the output layer outputs a historical trend reflecting the long-term performance change pattern. The sliding window technique effectively filters out the random fluctuations of single-time data through aggregation analysis of multiple sampling points, making the instantaneous degradation degree more closely match the short-term true performance state. LTM, with its advantage in processing long-period time-series data, solves the problem that traditional statistical methods struggle to capture long-term performance trends, achieving dual-dimensional coverage of performance characteristics—both short-term instantaneous and long-term trends—which is more comprehensive than single-dimensional extraction.

[0072] In redundant communication networks, the environmental interference risk of links depends not only on the strength of the interference source itself, but also on the position and movement status of the nodes at both ends of the link. For example, in industrial scenarios, the movement of equipment nodes can change the spatial orientation of the link, causing it to enter or leave the influence range of electromagnetic interference sources in the workshop. Conventional assessments only calculate the risk based on the strength of the interference source without considering the spatial information of the nodes, resulting in a disconnect between the risk probability and the actual scenario. To address this, we need to obtain information on the node position, movement trajectory, and interference source. The movement trajectory includes the movement speed and direction, and the interference source information includes the type of interference source, its installation location, and its influence range. We need to establish a two-dimensional topological coordinate system with the node as the origin, convert the node position into relative coordinates in the two-dimensional topological coordinate system, and combine the movement trajectory to generate a dynamic trajectory curve of the node in the two-dimensional topological coordinate system. This clarifies the real-time orientation of the link in the topological space and ultimately converts it into spatial topological coordinates.

[0073] This analysis examines the real-time trajectory of links within the topology space and the degree of overlap with the influence range of interference sources. For example, it determines whether the link is completely within the influence range of the interference source and the proportion of the overlapping segment length to the total link length. Combined with the link's transmission protocol type, its sensitivity to different interference sources is determined. For instance, wired links are less sensitive to electromagnetic interference than wireless links. Both factors are considered to determine the link's risk level from environmental interference, categorized as high, medium, and low, and then converted into corresponding risk probabilities. Spatial topology coordinate transformation is achieved through node information, enabling deep integration of node spatial information with the network topology without the need for additional equipment. This upgrades environmental interference assessment from single-source interference assessment to a precise assessment of the spatial relationship between the interference source and the link. Combining link sensitivity with risk probability optimization avoids unreasonable judgments of consistent risk for the same interference source across different links, significantly improving the scenario adaptability of risk probability.

[0074] S2. When the vulnerability index of any link exceeds the index threshold, topology reconstruction is triggered. Based on the current topology, highly vulnerable links are shielded, and alternative topologies that meet basic connectivity are explored in the redundant communication network using graph theory algorithms. The global robustness, service assurance, and reconstruction cost of each alternative topology are weighted and evaluated, and the target topology is selected based on the evaluation results.

[0075] Furthermore, shielding highly vulnerable links includes:

[0076] When the vulnerability index of any link exceeds the index threshold, topology reconstruction is triggered, and all marked highly vulnerable links are identified based on the current topology.

[0077] Associate the connection role and service priority of highly vulnerable links in the current topology of redundant communication networks, and determine the support weight of highly vulnerable links for services based on service priorities;

[0078] The impact coefficient of highly vulnerable links on the connectivity of the current topology is determined based on the connection role, and the shielding impact level of highly vulnerable links is divided based on the support weight and the impact coefficient.

[0079] Assess the topological and service impact of shielding highly vulnerable links at each shielding impact level, shield highly vulnerable links sequentially, and determine whether to suspend the shielding operation based on the connectivity of the current topology during the shielding process.

[0080] Exceeding the vulnerability index of links in industrial production workshops could be due to temporary fluctuations caused by instantaneous interference or continuous degradation caused by link aging. If topology reconfiguration is triggered solely by instantaneous fluctuations, production operations will be frequently interrupted. If continuously degraded links are overlooked, the risk of failure will increase. Furthermore, highly vulnerable links are scattered across different production units, requiring precise location of their physical positions and topology associations to avoid subsequent shielding operations affecting unrelated production units. The vulnerability index of each link is acquired at fixed intervals, with the fixed interval matching the minimum transmission cycle of workshop operations to ensure no critical fluctuations are missed. When the vulnerability index of a link first exceeds a preset threshold, reconfiguration is not immediately triggered. Instead, an observation period is initiated, continuously monitoring changes in the vulnerability index within this period. If the vulnerability index remains above the threshold without significant decline, it is considered continuous degradation, and topology reconfiguration is automatically triggered. If the vulnerability index falls below the threshold, it is considered instantaneous interference, and the triggering of topology reconfiguration is cancelled.

[0081] If the vulnerability index of a link exceeds the threshold and continues to deteriorate, the link is identified as a high-vulnerability link. The physical port identifiers of the nodes at both ends of the link are obtained through the LLDP protocol, and then associated with the production unit to which the physical port identifier belongs. Finally, the link identifier, physical port identifier, production unit, and vulnerability index are integrated into structured data and marked as high-vulnerability link information. The dual-trigger logic effectively filters out false reconstructions caused by momentary interference, reducing unnecessary interruptions to production operations. The structured data, combined with production unit and physical port information, makes the location of high-vulnerability links more consistent with the physical layout of the production workshop, providing a clear basis for subsequent priority shielding of services by unit.

[0082] Highly vulnerable links exhibit significant functional differences within the topology. Some links serve as core channels connecting the main controller and the server, and their disconnection would cause a global service interruption. Other links only connect edge sensors and area PLCs, and their disconnection would only affect local data acquisition. Furthermore, a single link may carry multiple types of services simultaneously. If all links are evaluated based on a single role or service priority, it could lead to biased decisions in subsequent shielding operations. Therefore, correlation analysis is needed to differentiate the actual importance of each link. A reachability traversal is performed on highly vulnerable links, simulating the disconnection of the link and traversing the connectivity status of all nodes to correlate the connection role of the highly vulnerable link in the current topology of the redundant communication network. If the disconnection results in a connectivity interruption between the main controller and the server, or between multiple production units, i.e., the topology splits into multiple independent subnets, then the link is marked as a backbone role. If only the connectivity between edge sensors and area PLCs is interrupted, and other core nodes are unaffected, then it is marked as a branch role.

[0083] Real-time traffic of highly vulnerable links is acquired, and service types are distinguished from the acquired data packets based on protocol fields. The proportion of data packets for different services is statistically analyzed, and combined with preset service priority rules, control commands have higher service priority than status data. The service support weight of the highly vulnerable link is calculated by weighting the proportion of data packets. For example, if a link carries 60% control commands and 40% status data, its support weight is more biased towards the higher priority of control commands. In this way, the connection role is determined by reachability traversal, avoiding the bias of subjective judgment based solely on link location. The connection role division is more in line with the actual function of the topology. The support weight is calculated according to the service proportion to ensure that the support weight can truly reflect the degree of support of the link for services of different priorities, providing a quantitative basis for subsequent shielding layer division.

[0084] The impact of link shielding depends not only on the connection role and business support weight, but also on the workshop production rhythm. For example, during peak production periods, although a branch link may be a branch role, shielding it can lead to the interruption of industrial processes and affect the overall production progress. Therefore, the activity status of production units needs to be included in the impact assessment to avoid the disconnect between technical indicators and actual production needs. The connection role and support weight are integrated into the basic impact value, that is, the backbone role corresponds to a higher basic impact value, and the higher the support weight, the higher the basic impact value. Then, the current production plan is retrieved from the production system to identify production units in an active state, such as production units that are performing industrial tasks. A production adaptation coefficient is added to the high-vulnerability links of production units in an active state. The production adaptation coefficient of links of production units in an active state is higher than that of links of production units in an inactive state. The basic impact value and the production adaptation coefficient are superimposed to obtain the final impact coefficient.

[0085] The impact levels are determined by ranking the impact coefficients. Links with the highest impact coefficients are classified as high-impact, requiring final blocking to avoid immediate disruption of core production. These high-impact links include backbone links in active production units supporting high-priority business. Links with medium impact coefficients are classified as medium-impact, including backbone links in inactive production units or high-weight branch links in active production units. Links with the lowest impact coefficients are classified as low-impact, requiring priority blocking to minimize production disruption. These low-impact links include low-weight branch links in inactive production units. A production adaptation coefficient is introduced to expand impact assessment from a purely technical dimension to a dual technical and production dimension, ensuring that the blocking impact level classification aligns with the actual production rhythm in the workshop. Ranking the blocking impact levels by impact coefficient clarifies the priority of blocking operations, preventing core business interruptions due to improper order.

[0086] If the load and connectivity of remaining links are not monitored when blocking links in order of impact level, it can lead to a surge in vulnerability index on remaining links due to a sudden increase in load after blocking low-impact links, triggering a new topology reconfiguration. It can also cause a sudden interruption of connectivity between core nodes after blocking medium-impact links, making it impossible to support temporary service transmissions. Therefore, a real-time monitoring and pause mechanism needs to be added during the blocking process to balance blocking progress and network stability. Blocking operations should be performed in the order of low-impact, medium-impact, and high-impact levels. Before each blocking operation, it should be determined which remaining links the traffic carried by the high-vulnerability link will be transferred to after blocking, and whether the load on these links will exceed the safety threshold. If it does, blocking should be suspended, and traffic allocation on the remaining links should be optimized first.

[0087] After the blocking operation is executed, the connectivity of core nodes is immediately checked to ensure topology reachability between core nodes, and the online rate of edge nodes is counted to ensure that most edge nodes can still upload data normally. If the connectivity of core nodes is interrupted, all blocking operations are immediately suspended and the most recently blocked link is restored. If the online rate of edge nodes drops significantly, i.e., below the normal level, only high-impact blocking is suspended, while the low / medium-impact links that have been blocked are retained to avoid further impact on connectivity. Load assessment before blocking avoids secondary risks caused by traffic shifting, and connectivity monitoring after blocking ensures that core services are not interrupted. Segmented suspension ensures the efficiency of removing high-risk links and maximizes the temporary communication capability of redundant communication networks, reserving a stable network environment for subsequent alternative topology exploration.

[0088] Specifically, such as Figure 2 As shown, the alternative topologies explored include:

[0089] Based on the remaining links after shielding highly vulnerable links, and combined with the backup links and node resources of redundant communication networks, a topology resource pool is formed, labeled with link bandwidth, anti-interference capability and transmission delay.

[0090] Node groups are divided according to the service interaction characteristics of nodes in redundant communication networks, joint constraints are set for links between different node groups, and joint constraints are determined based on the labeling information of the topology resource pool.

[0091] Based on joint constraints, a graph theory algorithm with link redundancy constraints is used to explore paths in the topology resource pool, generate a basic topology that satisfies basic connectivity, and supplement redundant paths with adaptive annotation information to expand the topology and generate a set of candidate topologies.

[0092] Calculate the betweenness centrality of each link in each candidate topology, remove redundant links whose betweenness centrality is less than the betweenness threshold and do not satisfy the joint constraints, and obtain the candidate topology structure.

[0093] After shielding highly vulnerable links, the remaining links will exhibit performance differences, such as insufficient bandwidth or weak anti-interference capabilities in some links. Furthermore, backup links may be dormant. If directly used for alternative topology exploration, they will generate topologies that cannot meet business requirements. Additionally, the lack of labeling of key link attributes will lead to a lack of basis for subsequent constraint settings. Based on the remaining links after shielding highly vulnerable links, where the remaining links have been confirmed to be connected and under normal load, inactive backup links, including redundant wireless links and backup fiber optic links, are retrieved. The physical connectivity of backup links is verified by sending LLDP probe frames. Links that do not respond to multiple probe frames are determined to be unusable and removed from the backup link. Available backup links are activated and added to the topology resource pool.

[0094] The system acquires real-time port traffic and calculates the current available bandwidth, which is the link bandwidth obtained by subtracting the currently used bandwidth from the nominal link bandwidth. It then retrieves the link type, recognizing that Ethernet links have higher interference immunity than wireless links, and combines this with historical interference immunity records, such as the number of historical failures caused by interference, to comprehensively determine the interference immunity level and obtain the interference immunity capability. The system tests the round-trip time between the nodes at both ends of the link, taking the average of multiple tests to obtain the transmission latency. Simultaneously, it supplements the active status and historical failure records. Activation status includes both already activated and newly activated, and historical failure records include the number of recent failures, forming a topology resource pool labeled with link bandwidth, interference immunity capability, and transmission latency. Connectivity verification of backup links ensures that all links in the topology resource pool are available, avoiding the selection of unavailable links during subsequent exploration. Attribute labeling provides accurate basis for subsequent constraint setting, making the generated alternative topologies more aligned with business needs and reducing the generation of invalid topologies.

[0095] The business interactions of nodes in the production workshop exhibit significant unit clustering, meaning that interactions between nodes within the same production unit are frequent, while interactions between nodes across units are less frequent. Ignoring this clustering would result in a topology containing numerous cross-unit links, leading to uneven link load and susceptibility to interference. Furthermore, different production units have varying performance requirements for links, necessitating targeted constraints to avoid resource waste or performance deficiencies caused by a one-size-fits-all approach. The solution involves first obtaining currently active production units and their corresponding nodes from the production system, and then grouping nodes within these active production units. Next, the interaction frequency between these nodes is determined by counting the number of data packets sent between nodes within a certain period. Nodes with high interaction frequencies are grouped together, ultimately forming node groups corresponding to production units. Simultaneously, the core node is retained as a shared node, belonging to all node groups to ensure that each node group can communicate with the core node.

[0096] When setting joint constraints for links between different node groups, the annotation information of the topology resource pool and the service requirements of each node group are combined. For node groups carrying control commands, joint constraints of low latency and high anti-interference are set, where the link transmission latency must meet the real-time requirements of the control commands, and the anti-interference level must match the electromagnetic environment of the workshop. For node groups carrying status data, joint constraints of sufficient bandwidth and low cost are set, where the link bandwidth must meet the data transmission volume, and activated links are selected first to reduce activation costs. For cross-group links, constraints of media diversity are added to avoid the use of the same type of media for primary and backup links. For example, the primary link uses fiber optic and the backup link uses wireless, which improves anti-interference redundancy. Node groups are divided according to production units and interaction characteristics to make the topology structure highly matched with the distribution of business flows, reduce unnecessary cross-group data transmission, reduce link load, and ensure that the service requirements of different node groups are met. This avoids performance deficiencies or resource waste caused by single constraints, while media diversity constraints improve the anti-interference capability of the topology structure.

[0097] Generating only a single basic topology can lead to single-point risks. For example, if the basic topology depends on a critical link, and that link fails, the topology becomes invalid and cannot cope with weight changes in subsequent evaluations. For instance, if the weight of a certain type of constraint is increased, the original basic topology may no longer be optimal. At the same time, the basic topology must prioritize the connectivity of core nodes to avoid impacting core business due to edge nodes. First, a core-first graph theory algorithm is used to use core nodes as the initial node set of the topology. Links that satisfy joint constraints and can connect core nodes to each node group are selected from the topology resource pool to ensure that core nodes are connected to all node groups. Then, links connecting nodes within node groups are gradually added to eventually form a basic topology that satisfies the connectivity of core nodes and the basic connectivity of nodes within each node group.

[0098] For critical links in the basic topology, redundant links that meet the following conditions are selected from the topology resource pool: different media type from the critical link, able to connect the same node pairs, and not used by the basic topology. Different media types are to comply with media diversity constraints, same connections ensure redundancy, and unused links avoid resource duplication. One to two redundant links are configured for each critical link. By replacing different combinations of redundant links, redundant paths adapted to the annotation information are supplemented to expand the topology, generating multiple alternative topologies with different structures, forming a candidate topology set. A core-first graph theory algorithm ensures that the basic topology prioritizes the connectivity of core services, avoiding the impact of edge nodes on the overall system. The candidate topology set generated by combining redundant links provides diverse options for subsequent evaluation, avoiding the limitations of a single topology. Simultaneously, the selection and expansion of redundant links ensures the anti-interference capability and resource utilization of the candidate topologies.

[0099] Some redundant links in the candidate topology set may be kept in reserve and not used, meaning they are only activated in case of extreme failures and have no traffic transmission in daily operations. These links not only consume resources but also increase the complexity of the topology. At the same time, some redundant links may have failure correlations with other links. For example, if two links are deployed on the same physical pipeline, they will fail simultaneously when the pipeline fails. Retaining such links cannot improve the actual risk resistance capability. Therefore, invalid redundant links need to be removed. First, the daily operation status of the candidate topology is simulated, and the proportion of business traffic carried by each link is counted. Links with a high proportion of business traffic are assigned higher betweenness weights. Finally, the betweenness centrality of each link is obtained to reflect the actual importance of the link in daily operation.

[0100] If the betweenness centrality of a redundant link is less than the betweenness threshold, it indicates that it is rarely used in daily operations and has a fault correlation with other links in the candidate topology. For example, if the deployment information shows that two links share the same physical pipe or power supply module, it is determined to be an invalid redundant link and removed from the candidate topology. If the betweenness centrality is less than the betweenness threshold but has no fault correlation, such as an independently deployed link that can still serve as a backup in the event of an extreme failure, it is retained as an emergency redundant link and is not removed. After performing the above operations on all candidate topologies, the candidate topology structure is obtained. Betweenness centrality avoids the bias caused by calculating betweenness solely based on the topology. For example, a link with a high betweenness centrality may be misjudged as an important link if it has no traffic in daily operations. The dual removal criteria ensure that truly invalid redundant links are removed, while emergency redundant links are retained. This simplifies the topology complexity and reduces resource consumption without sacrificing the ability to withstand risks under extreme failures.

[0101] Specifically, selecting the target topology includes:

[0102] Based on the annotation information of the topology resource pool, the backup links and node resources of the redundant communication network, and the service priority, the evaluation indicators including global robustness, service assurance and reconstruction cost are quantified.

[0103] The weights of the evaluation indicators are adjusted based on the proportion of redundant links and the proportion of spare resources in the alternative topology, forming an indicator weight matrix.

[0104] Simulate the random removal of non-highly vulnerable links from alternative topologies, calculate the decrease in connectivity retention rate and service assurance of alternative topologies after removal, and determine whether to add a risk resistance index.

[0105] The basic evaluation value of the candidate topology is obtained by weighting the index weight matrix, and the comprehensive evaluation value is obtained by superimposing the risk resistance index. The target topology is then selected based on the comprehensive evaluation value.

[0106] The merits of alternative topologies require a comprehensive evaluation across multiple dimensions. Focusing solely on global robustness may lead to the selection of topologies with excessive redundancy or high costs, while focusing solely on reconstruction costs may result in topologies with insufficient robustness or susceptibility to failure. Furthermore, the evaluation must be based on objective data to avoid subjective judgment. Global robustness is represented by the product of the proportion of redundant links and the proportion of high-interference-resistance links in the alternative topology. Service reliability is represented by the bandwidth and latency compliance rates of links corresponding to high-priority services. Reconstruction cost is represented by the ratio of the number of spare resources to be activated to the reuse rate of remaining links. This multi-dimensional quantification avoids the bias of a single indicator evaluation, allowing the merits of alternative topologies to be compared from different perspectives. Quantification based on objective data ensures the fairness of the evaluation results and reduces the influence of subjective factors on the selection decision.

[0107] The initial weights of evaluation metrics are mostly based on business priorities, but they do not consider the resource characteristics of the topology itself, which directly affects the actual value of the evaluation metrics. For example, if a candidate topology has a high proportion of redundant links, its global robustness is theoretically stronger, but too many redundant links will increase maintenance costs. If the weight of high robustness is still used, it will lead to the selection of a high-cost and low-efficiency topology. Another candidate topology has a high proportion of spare resources, and the actual reconstruction cost is higher. If the weight of the reconstruction cost metric is not adjusted, its long-term operation and maintenance burden may be underestimated. For each candidate topology, the relationship between the number of redundant links and the total number of links is calculated to obtain the proportion of redundant links, and the relationship between the number of activated spare resources and the total resources of the topology is calculated to obtain the proportion of spare resources.

[0108] The proportion of redundant links is positively correlated with global robustness, as a higher proportion theoretically provides stronger fault tolerance. However, it is also positively correlated with reconstruction costs, as redundant links require additional maintenance. The proportion of spare resources is positively correlated with reconstruction costs, as a higher proportion leads to higher configuration and authentication costs for activating new resources. If the spare resources are anti-interference media, it is positively correlated with service assurance, thus improving the stability of high-priority service transmission. The weights of evaluation indicators are adjusted based on the proportion of redundant links and spare resources in the alternative topologies. If a candidate topology has a high proportion of redundant links, the weight of the global robustness indicator is appropriately increased to acknowledge its fault tolerance advantage, while the weight of the reconstruction cost indicator is simultaneously increased to balance the maintenance burden caused by redundancy and avoid excessive preference for highly redundant topologies.

[0109] If the proportion of spare resources in the alternative topology is high, the weight of the reconstruction cost indicator is significantly increased to highlight the cost impact of activating new resources. If the spare resources are anti-interference type, the weight of the business assurance indicator is slightly increased to acknowledge their support value for the business. If both proportions are low, the weight of the business assurance indicator is increased to prioritize the actual support capability of existing resources for the business and the weight of global robustness is reduced because the robustness advantage is limited due to insufficient redundancy. The adjusted weights of global robustness, business assurance, and reconstruction cost are integrated to ensure that the sum of the weights reflects the relative importance of each indicator. For example, in a high-redundancy topology, the weights of global robustness, reconstruction cost, and business assurance are balanced, while in a low-spare-resource topology, the weight of business assurance is dominant, forming an indicator weight matrix that adapts to the characteristics of the topology resources.

[0110] By adjusting the weights of redundant links and spare resources, the weight matrix of indicators is upgraded from a static setting that relies solely on business priority to a dynamic adaptation driven by both business needs and resource characteristics. This avoids evaluation bias caused by ignoring the actual state of topology resources. For example, it avoids choosing a high-cost topology with excessive redundancy in pursuit of global robustness, and avoids choosing a topology that is difficult to maintain in the long term due to underestimating the cost of spare resources. This ensures that the evaluation is more in line with the actual carrying capacity of redundant communication network resources.

[0111] Risk assessments of alternative topologies often focus on failure scenarios of highly vulnerable links, neglecting random failures of non-highly vulnerable links. Although these failures have a low probability, they can lead to a break in topology connectivity or a sharp drop in service availability, resulting in a one-sided assessment of topology resilience and failing to ensure its stability under real-world complex failure scenarios. Non-highly vulnerable links are selected from the alternative topologies to determine the scope of links to be simulated for removal. Highly vulnerable links that have already been marked should be excluded, focusing on links that are prone to random failures during daily operation to avoid repeatedly assessing failure scenarios of highly vulnerable links. Based on the link's connection role in the current topology, non-highly vulnerable links with different connection roles are randomly selected for simulated removal to ensure coverage of both backbone and branch links, avoiding a one-sided scenario due to simulating only a single role link. The removal operation is implemented using a topology simulation tool, without changing the actual link status; only the link is marked as unavailable in the simulation environment.

[0112] In a simulation environment, a node reachability traversal algorithm is used to determine the connectivity status of all nodes in the candidate topology after removing non-highly vulnerable links. The focus is on checking whether all core nodes remain connected and whether edge nodes can connect to core nodes or their respective node groups. The relationship between the number of connected node pairs and the total number of node pairs before removal is calculated to obtain the connectivity retention rate. Core node connectivity is the key criterion, and ensuring core nodes remain uninterrupted is a priority. Then, the transmission quality of each service in the candidate topology is compared before and after removing non-highly vulnerable links. For high-priority services, latency and interruption frequency are the primary concerns, while for low-priority services, packet loss rate and data integrity are the primary concerns. The system analyzes whether the service transmission quality still meets the preset requirements after removal. If the transmission quality deteriorates, for example, if transmission latency increases or the data packet loss rate rises, the magnitude of the deterioration is calculated, i.e., the attenuation value of the service guarantee. If there is no deterioration or the requirements are still met, the attenuation value is considered low.

[0113] The decision to add a risk resistance index is based on a combination of connectivity retention rate and service assurance attenuation value. If all core nodes remain connected and edge nodes have high connectivity rates, and all services have low assurance attenuation values ​​(meaning transmission quality still meets requirements), then the candidate topology is considered to have strong resistance to random failures, and a risk resistance index is added. If core nodes experience connectivity interruptions, or high-priority services have high assurance attenuation values ​​(meaning real-time requirements cannot be met), then the risk resistance is considered weak, and no risk resistance index is added. If only edge nodes have slightly low connectivity rates and low-priority services have low attenuation values, then a lower risk resistance index is added. This supplements the assessment scenario for random failures of non-highly vulnerable links, avoids the one-sidedness of risk resistance assessments that focus solely on high-vulnerability link failures, and makes the assessment more closely reflect the complex scenarios in real-world networks where multiple failure types coexist.

[0114] By simulating the removal of backbone and branch links, combining connectivity assessments of core and edge nodes, and analyzing service-level reliability degradation, the risk assessment ensures that it not only focuses on topology stability but also considers service transmission quality, avoiding the hidden danger of service interruption despite topology connectivity. The differentiated addition of the risk resistance index provides a quantitative basis for the calculation of subsequent comprehensive assessment values, enabling the selection of target topology to not only consider normal operation performance but also prioritize topologies with strong resistance to random failures, thereby improving the long-term reliability of redundant communication networks.

[0115] The comprehensive evaluation value needs to integrate the basic evaluation value and the risk resistance index to fully reflect the normal performance and risk resistance capability of the topology. At the same time, if the comprehensive evaluation values ​​of multiple candidate topologies are close, further screening is required through additional adaptability judgment to avoid subsequent operation and maintenance difficulties caused by selecting based solely on numerical values. The basic evaluation value of the candidate topology structure is calculated by weighting according to the indicator weight matrix. The basic evaluation value is then superimposed with the risk resistance index to obtain the comprehensive evaluation value of each candidate topology structure. Then, candidate topologies with significantly low comprehensive evaluation values ​​are first screened out. Then, the remaining candidate topologies are checked for weaknesses. If a certain indicator of a candidate topology structure is much lower than that of other candidate topologies, it is determined to have a fatal weakness and is excluded to avoid business problems caused by weaknesses in subsequent use.

[0116] If there are similar comprehensive evaluation values ​​among the remaining candidate topologies, historical adaptability and ease of maintenance are introduced as supplementary judgment criteria. Historical adaptability refers to the similarity of the candidate topology to the topologies that have performed well in the past in the production workshop, including the proportion of redundant links and the distribution of media types. Ease of maintenance refers to whether the links of the candidate topology are easy to monitor and maintain, such as whether they are centrally deployed or whether mature maintenance tools are used. The candidate topology with high historical adaptability and good ease of maintenance is selected as the target topology. Shortcoming check avoids the selection of candidate topologies with high comprehensive evaluation values ​​but fatal shortcomings, ensuring that the target topology has no obvious defects. The supplementary judgment criteria solve the selection problem when comprehensive evaluation values ​​are similar, so that the target topology not only has excellent current performance, but also matches the historical operating experience of the production workshop and is easy to maintain in the future, reducing long-term maintenance costs.

[0117] S3. Transmit the target topology path corresponding to the target topology structure in parallel with the current topology path. During the parallel transmission process, migrate the data stream of the current topology path to the target topology path in stages according to the service priority. After migration, release the link resources of the current topology path and monitor the operation feedback data of the target topology path to update the output vulnerability index and weighted evaluation parameters.

[0118] Furthermore, such as Figure 3 As shown, parallel transmission includes:

[0119] Establish parallel transmission channels between the target topology path and the current topology path, allocate transmission resources based on service priority, and initiate data flow migration of the current topology path in stages according to service priority;

[0120] Real-time monitoring and comparison of the path operation parameters of the two paths, and adjustment of the migration ratio of the data flow based on the degree of degradation of the target topology path relative to the current topology path;

[0121] Once the data flow of high-priority services has been fully migrated to the target topology path and the data flow of low-priority services has met the preset migration ratio, the link resources of the current topology path will be released in batches according to the connection role of the link.

[0122] Directly switching services from the current topology path to the target topology path could lead to service interruptions because the target topology path has not undergone stability verification. In particular, interruptions of high-priority control commands could cause production accidents. Furthermore, services with the same priority have different transmission resource requirements; uniform resource allocation could result in low-priority services preempting high-priority services. Therefore, it is necessary to establish parallel channels and allocate resources according to rules, and to migrate in stages. Parallel transmission channels should be established between the target and current topologies, with independent identifiers configured for each. The current topology path retains its original identifier, while the target topology path creates a new, dedicated identifier to avoid data packet conflicts at the link layer. Transmission resources should be allocated based on service priority to ensure stable transmission latency for high-priority services and to allow low-priority services to dynamically occupy idle resources when high-priority services have no transmission needs.

[0123] Then, data flow migration of the current topology path is initiated in stages according to business priority. Low-priority services are migrated first, with a small proportion of their data flow switched to the target topology path. The integrity of data packets on the target topology path is verified, and after confirming no packet loss or latency fluctuations, the migration ratio of data flow is gradually increased. After the migration of low-priority services is completed and the operation is stable, the migration of high-priority services is initiated. During the migration process, a hot backup of the current topology path is maintained, that is, the transmission of the current topology path is not interrupted. Independent identification avoids data packet conflicts between the two paths, and the allocation of transmission resources ensures the exclusive use of resources for high-priority services. Starting the migration in stages from low-priority services can verify the stability of the target topology path without affecting core production, and significantly reduce the risk of direct switching.

[0124] During parallel transmission, the performance of the target topology path fluctuates due to changes in the external environment. Maintaining a fixed migration ratio would increase the transmission latency of high-priority services, while the load on the current topology path would decrease with the migration ratio. Failure to adjust in time would result in resource waste. Therefore, it is necessary to monitor and dynamically adjust the migration ratio in real time. The path operating parameters of the two paths, including transmission latency, data packet loss rate, and link load, should be monitored and compared in real time, with the monitoring cycle synchronized with the service transmission cycle. Then, using the path operating parameters of the current topology path as a benchmark, the degree of degradation of the target topology path should be determined. If the transmission latency of the target topology path is higher than that of the current path and the data packet loss rate increases, it is considered degraded, and the migration ratio of high-priority services should be immediately reduced, and further migration should be suspended. If the path operating parameters of the target topology path are consistently better than those of the current topology path, the migration ratio should be gradually increased until all high-priority services are fully migrated. Real-time monitoring ensures a rapid response to path performance fluctuations, avoiding a decline in service transmission quality due to performance degradation. Dynamically adjusting the migration ratio ensures the stability of high-priority services while efficiently utilizing resources when the target topology path has good performance, balancing stability and resource utilization.

[0125] If all link resources of the current topology path are released at once after the migration is completed, there will be no backup path available due to sudden failure of the target topology path. At the same time, the links of the current topology path can be divided into core links and non-core links according to their connection roles. Core links carry critical business backups. If core links are released first, it will increase the risk of failure. Therefore, they need to be released in batches according to connection roles. When high-priority services have been fully transmitted in the target topology path and there are no abnormalities for 5 consecutive monitoring periods, and the migration ratio of low-priority services has reached the preset migration ratio, the release is carried out in batches in the order of non-core links to core links. The first batch of non-core links is released by sending a link disable command through the link layer protocol. After disabling, it is monitored whether the target topology path experiences a sudden increase in load due to the release of non-core links. If there are no abnormalities, the next stage is entered.

[0126] The second batch of core links is released. Before release, the load change of the target topology path after the core link is released is judged. The disable command is sent only after confirming that there is no overload risk. After each core link is released, the connectivity and service transmission quality of the target path are verified at one monitoring cycle. The next link is released only after ensuring that there are no abnormalities. The phased release avoids the risk of no backup caused by a one-time release. The priority release of non-core links can gradually verify the load carrying capacity of the target topology path. Simulation judgment and interval verification are performed before the core links are released to reduce the probability of overload or failure of the target topology path during the release process and ensure the safety of resource release.

[0127] Specifically, updating the output vulnerability index includes:

[0128] After releasing the link resources of the current topology path, monitor the operation feedback data of the target topology path. The operation feedback data includes path operation parameters and service migration quality.

[0129] The anti-interference performance of the target topology path's path operation parameters is correlated and compared with the sensitivity of the link to environmental risk scenarios in order to adjust the environmental correction factor.

[0130] Degradation trends are extracted based on the time-series changes of path operation parameters and similarity matching is performed with historical fault data. The correlation strength coefficient between environmental risk scenarios and performance degradation is adjusted according to the matching results.

[0131] Based on the service migration quality of different services during parallel transmission, the fusion weights of performance correction features and environmental correction factors are adjusted accordingly.

[0132] Based on the adjusted environmental correction factor, correlation strength coefficient, and fusion weight, nonlinear fusion is re-executed to dynamically calculate and update the output vulnerability index.

[0133] The vulnerability index is derived from historical link failure data and initial environment data, but it does not include the actual performance and service adaptability of the target topology path during operation. If only historical data is relied upon, the vulnerability index will fail to reflect the current true state of the link. Therefore, it is necessary to monitor the operational feedback data of the target topology path as a basis for correction. The operational feedback data includes path operation parameters and service migration quality. The path operation parameters include the anti-interference performance of the target topology path and the time-series performance degradation trend. The service migration quality includes the number of migration interruptions for services of different priorities and the changes in transmission quality of services after migration. The operational feedback data is acquired in real time and stored according to link identifiers. Each link of the target topology path corresponds to a set of independent operational feedback data to avoid data confusion between links. At the same time, the acquired operational feedback data is preprocessed to ensure data availability. Monitoring the operational feedback data fills the gap between historical data and actual operating status, providing a real and real-time basis for the correction of the vulnerability index. The classification and preprocessing according to link identifiers ensures that subsequent parameter adjustments can accurately correspond to each link and avoid correction deviations.

[0134] The environmental correction factor is based on the sensitivity of the link to environmental risk scenarios. However, the actual anti-interference performance of the target topology path in operation may not match the theoretical sensitivity. If the original environmental correction factor is still used, the vulnerability index may overestimate or underestimate the environmental risk of the link. Therefore, it is necessary to adjust it in combination with the actual anti-interference performance. The sensitivity of the corresponding link to environmental risk scenarios is retrieved from historical fault data, and the actual anti-interference performance of the link is extracted from the operation feedback data. When comparing the correlation, if the link sensitivity is high but the actual anti-interference performance is good, it is a high sensitivity level. However, if there is no obvious performance fluctuation in actual operation, the high value environmental correction factor is adjusted to the median value, indicating that the actual impact of the environment on the link performance degradation is less than expected.

[0135] If the sensitivity level is low but the actual anti-interference performance is poor, for example, it is classified as low sensitivity level, meaning it is theoretically insensitive to temperature, but the current environment indicates a sudden increase in link load. In this case, the low-value environmental correction factor is adjusted to the medium value, that is, it is adjusted from the low value to the medium value in steps of 0.3. Insensitivity to temperature is the low sensitivity level (low sensitivity level). A sudden increase in link load means that the increase in link load per unit time exceeds the normal operating baseline range, indicating that it should be low sensitivity level, but a sudden increase in link load actually occurs. If the sensitivity level is consistent with the actual anti-interference performance, the environmental correction factor remains unchanged. During the adjustment process, the historical adjustment records of similar links are referenced to ensure that the adjustment range conforms to the link characteristics. The environmental correction factor is adjusted based on the actual anti-interference performance, so that the environmental correction factor is upgraded from the theoretical value to a precise value that combines theory and reality, avoiding misjudgment of environmental risks by the vulnerability index and improving the scenario adaptability of the vulnerability index.

[0136] The correlation strength coefficient is established based on historical fault data. However, the current performance degradation trend of the target topology path may differ from the degradation pattern in the historical fault data. For example, the current load increase rate may be slower than the increase rate before the historical fault. If the original correlation strength coefficient is still used, the vulnerability index will not be able to reflect the true correlation between the current degradation trend and the environmental risk scenario. The performance degradation trend is extracted from the path operation parameters of the target topology path. That is, the temporal change pattern of the path operation parameters of the link is first sorted out to form the current degradation trend characteristics. The degradation trend data before the fault of similar links is retrieved from the historical fault database, and the similarity between the current degradation trend and the historical trend is calculated by using the cosine similarity algorithm.

[0137] If the similarity is high, meaning the current degradation trend highly overlaps with the pre-failure trend, the correlation strength coefficient is increased to indicate that the probability of performance degradation caused by the current environmental risk scenario is close to the historical failure probability. If the similarity is low, meaning the current trend is more gradual and there are no sudden features before the failure, the correlation strength coefficient is decreased. If there is no similar historical failure data, the correlation strength coefficient is adjusted based on the design life and usage duration of the link. For example, if the usage duration is close to the design life, the correlation strength coefficient is appropriately increased. By matching the degradation trend with historical failure data, the correlation strength coefficient is transformed from a static historical value to a value that adapts to the current degradation trend, ensuring that the assessment of the correlation between environmental risk scenarios and performance degradation is more in line with the current reality and avoiding misjudgment of failure risk by the vulnerability index.

[0138] The fusion weight is set based on business priority, but the business migration quality of the target topology path reflects the actual impact of performance and environmental factors on the business. If the business migration quality of a high-priority business is poor, i.e., it is frequently interrupted, it indicates that the fusion weight of the performance correction feature is insufficient, and the fusion weight needs to be adjusted to prioritize performance. Therefore, the fusion weight needs to be adjusted according to the difference in business migration quality. The business migration quality of different priority businesses is extracted from the operation feedback data. For high-priority businesses, the focus is on the number of migration interruptions and latency fluctuations after migration, while for low-priority businesses, the focus is on the packet loss rate and data update integrity after migration.

[0139] When adjusting for differences, if the migration quality of high-priority services is poor (i.e., multiple interruptions occur), the proportion of performance correction features in the fusion weight is increased to prioritize improving service quality through performance optimization and reduce the proportion of environmental correction factors. If the migration quality of low-priority services is poor (i.e., high data packet loss rate) and analysis reveals it to be related to environmental interference, such as strong electromagnetic interference during the migration period, the weight of environmental correction factors is increased. If the migration quality of all services is excellent, the fusion weight remains unchanged, only slightly adjusted to match the current service load. Adjusting the fusion weight based on the differences in service migration quality transforms the fusion weight from a static setting based on priority to a dynamic adaptation based on actual service performance. This ensures that the vulnerability index calculation prioritizes reflecting the core factors affecting service quality and improves the vulnerability index's alignment with service needs.

[0140] The environmental correction factor, correlation strength coefficient, and fusion weights have all been adjusted based on the actual operating data of the target topology path. The original vulnerability index can no longer reflect the current true vulnerability status of the link. If the original vulnerability index is continued to be used, it will lead to misjudgment of the timing of the next topology reconstruction. The aforementioned nonlinear fusion model is called, with the adjusted environmental correction factor and correlation strength coefficient as model inputs, and the fusion weights as the connection weights from the input layer to the hidden layer. The hidden layer processes the nonlinear coupling relationship between the performance correction feature and the environmental correction factor through the activation function, and the output layer re-outputs the updated vulnerability index.

[0141] Then, the new vulnerability index is verified against the actual operating status of the target topology path. If the vulnerability index is high and there are signs of performance degradation in actual operation, or if the vulnerability index is low and the actual operation is stable, the verification passes. If the verification fails, the previous parameter adjustment process is checked back, the deviation is corrected, and the index is recalculated. Finally, the verified vulnerability index is stored according to the link identifier, replacing the original vulnerability index. The recalculation and verification ensure that the updated vulnerability index can accurately reflect the current comprehensive vulnerability status of the link, avoiding the distortion of the vulnerability index caused by outdated parameters. Consistency verification reduces calculation errors and improves the reliability of the vulnerability index, so as to ensure timely triggering of topology reconstruction and guarantee stable network operation.

[0142] Specifically, the parameters for updating the weighted evaluation include:

[0143] The link load fluctuation and anti-interference performance of the path operation parameters are correlated with global robustness; the remaining link reuse rate of the path operation parameters is correlated with reconstruction costs; and the quality of service migration is correlated with service assurance.

[0144] The evaluation metrics of the target topology during the selection phase are compared with the corresponding path operation parameters and service migration quality during the parallel transmission phase to obtain the deviation coefficient of the evaluation metrics.

[0145] The adjustment priority of the indicator weights is determined based on the magnitude of the deviation coefficient, and the indicator weights are adjusted in combination with the deviation coefficient to regenerate the indicator weight matrix for target topology selection.

[0146] The weighted evaluation parameters are set based on the static characteristics of the topology and do not take into account the actual performance of the target topology path in operation. If the static evaluation value of a certain evaluation indicator differs from the actual performance, it will lead to evaluation distortion when selecting the topology structure in the future. The correlation between the operational feedback data is established according to the type of evaluation indicator. The link load fluctuation and anti-interference performance of the path operation parameters in the target topology path are correlated with the global robustness. The link load fluctuation is small and the anti-interference performance is good, which indicates that the actual global robustness is high. The remaining link reuse rate in the path operation parameters is correlated with the reconstruction cost. The high remaining link reuse rate indicates that fewer new resources are used and the cost is low during actual reconstruction. The remaining link reuse rate refers to the resource utilization efficiency of the target topology path on the current remaining links.

[0147] The quality of business migration is linked to business assurance. Fewer business migration interruptions and better transmission quality indicate a high level of actual business assurance. During the linkage process, data tagging is used to ensure that each data point accurately corresponds to the target evaluation indicator, avoiding confusion in the linkage. The linkage between operational feedback data and evaluation indicators establishes a bridge between static evaluation and actual performance, so that subsequent parameter adjustments no longer depend on static topology features, but are based on real operational results, thus improving the practicality of evaluation indicator parameters.

[0148] When selecting a target topology, the assessment of global robustness, service assurance, and reconstruction cost is based on the simulated values ​​of the topology. The actual operating data of the target topology path represents the real values. If the simulated values ​​differ significantly from the real values—meaning high global robustness in simulation but frequent fluctuations in actual load—it indicates poor adaptability of the original weighted evaluation parameters. A deviation coefficient is needed to quantify the difference and provide direction for parameter adjustments. The evaluation indicators of the target topology during the selection phase are compared with the corresponding path operating parameters and service migration quality during the parallel transmission phase. The direction and magnitude of the difference between the simulated and actual values ​​are analyzed. If the simulated value is higher than the actual value, it is marked as a positive deviation; if the simulated value is lower than the actual value, it is marked as a negative deviation. The magnitude of the deviation is quantified through qualitative description, forming a deviation coefficient for each evaluation indicator. The deviation coefficient clearly identifies the difference between the original evaluation indicator and the actual performance, avoiding blind parameter adjustments. By analyzing the direction and magnitude of the deviation, it is possible to accurately determine which evaluation indicator's weight needs adjustment and in what direction, improving the targeting of parameter adjustments.

[0149] Different evaluation metrics have varying impacts on topology selection. Significant deviations in global robustness can lead to topology selections failing to meet core business stability requirements, necessitating priority adjustment of their metric weights. Conversely, minor deviations in reconstruction cost have minimal impact on topology selection and can be adjusted later. Therefore, adjustment priorities must be determined based on deviation magnitude to avoid unresolved core issues due to average adjustments. The adjustment priority of metric weights is determined by the magnitude of the deviation coefficient, with significant deviations assigned as first-level priority, moderate deviations as second-level priority, and minor deviations as third-level priority. When adjusting metric weights, for first-level priority evaluation metrics, if the simulated global robustness value is higher than the actual value, it indicates that the original metric weight overestimated global robustness. In this case, the proportion of global robustness in the metric weight matrix should be reduced, while the weight of business assurance should be appropriately increased to ensure business stability.

[0150] For secondary priority evaluation indicators, if the simulated value of service availability is lower than the actual value, it indicates that the original indicator weight underestimates service availability, and its weight is slightly increased. For tertiary priority evaluation indicators, if the reconstruction cost deviation is minor, the indicator weight is only fine-tuned to match the current resource utilization. For example, when the remaining link reuse rate is high, the indicator weight of reconstruction cost is reduced. After the adjustment, a new indicator weight matrix is ​​generated to ensure that the sum of the weights of each indicator in the indicator weight matrix is ​​consistent with the original indicator weight matrix, that is, the total weight is kept at 1, to avoid confusion in the evaluation scale due to changes in the total weight of the indicator weights. Adjusting the indicator weights according to the adjustment priority ensures that core deviation problems are resolved first, avoiding waste of resources on minor deviations. The new indicator weight matrix is ​​generated based on the actual deviation and is more in line with the actual operation of redundant communication networks than the original indicator weight matrix, which can improve the evaluation accuracy of the next target topology selection.

Claims

1. A link self-healing topology control method for redundant communication networks, characterized in that, include: Acquire performance and environmental data of redundant communication networks, integrate the instantaneous degradation and historical trends of performance data with the risk probability of environmental data, dynamically calculate and output the vulnerability index of each link in the redundant communication network; When the vulnerability index of any link exceeds the index threshold, topology reconfiguration is triggered. Based on the current topology, highly vulnerable links are shielded, and alternative topologies that meet connectivity requirements are explored in the redundant communication network using graph theory algorithms. The global robustness, service assurance, and reconfiguration cost of each alternative topology are evaluated in a weighted manner, and the target topology is selected based on the evaluation results. The target topology path corresponding to the target topology structure is transmitted in parallel with the current topology path. During the parallel transmission process, the data stream of the current topology path is migrated to the target topology path in stages according to the service priority. After the migration, the link resources of the current topology path are released, and the operation feedback data of the target topology path is monitored to update the output vulnerability index and weighted evaluation parameters.

2. The link self-healing topology control method for redundant communication networks as described in claim 1, characterized in that, The output vulnerability index includes: The system acquires performance and environmental data for each link in a redundant communication network, extracts features from the performance and environmental data, and obtains instantaneous degradation, historical trends, and risk probabilities. By combining business load and normal operation benchmarks, the deviation between instantaneous degradation and historical trends is corrected to obtain performance correction characteristics that reflect link performance degradation. Based on the sensitivity of the link to environmental risk scenarios, the risk probability is converted into an environmental correction factor that affects performance degradation. Based on historical fault data, establish the correlation strength coefficient between environmental risk scenarios and performance degradation, and adjust the weight ratio of the correlation strength coefficient according to the service priority of the current redundant communication network to form a fusion weight of performance correction characteristics and environmental correction factors. The performance correction characteristics and environmental correction factors are nonlinearly fused according to the fusion weights to dynamically calculate and output the vulnerability index of each link in the redundant communication network.

3. The link self-healing topology control method for redundant communication networks as described in claim 2, characterized in that, The characteristics of the extracted performance data and environmental data include: The instantaneous degradation of performance data is extracted by a sliding time window, and the historical trend of performance data is extracted based on a long short-term memory network. The node locations and movement trajectories in the environmental data are converted into spatial topological coordinates, and the probability of the link being affected by environmental interference is calculated by combining the interference source information.

4. The link self-healing topology control method for redundant communication networks as described in claim 3, characterized in that, Exploring the alternative topologies includes: Based on the remaining links after shielding highly vulnerable links, and combined with the backup links and node resources of redundant communication networks, a topology resource pool is formed, labeled with link bandwidth, anti-interference capability and transmission delay. Node groups are divided according to the service interaction characteristics of nodes in redundant communication networks, joint constraints are set for links between different node groups, and joint constraints are determined based on the labeling information of the topology resource pool. Based on joint constraints, a graph theory algorithm with link redundancy constraints is used to explore paths in the topology resource pool, generate a basic topology that satisfies connectivity, and supplement redundant paths with adaptive annotation information to expand the topology and generate a set of candidate topologies. Calculate the betweenness centrality of each link in each candidate topology, remove redundant links whose betweenness centrality is less than the betweenness threshold and do not satisfy the joint constraints, and obtain the candidate topology structure.

5. The link self-healing topology control method for redundant communication networks as described in claim 4, characterized in that, The shielded highly vulnerable links include: When the vulnerability index of any link exceeds the index threshold, topology reconstruction is triggered, and all marked highly vulnerable links are identified based on the current topology. Associate the connection role and service priority of highly vulnerable links in the current topology of redundant communication networks, and determine the support weight of highly vulnerable links for services based on service priorities; The impact coefficient of highly vulnerable links on the connectivity of the current topology is determined based on the connection role, and the shielding impact level of highly vulnerable links is divided based on the support weight and the impact coefficient. Assess the topological and service impact of shielding highly vulnerable links at each shielding impact level, shield highly vulnerable links sequentially, and determine whether to suspend the shielding operation based on the connectivity of the current topology during the shielding process.

6. The link self-healing topology control method for redundant communication networks as described in claim 5, characterized in that, The selected target topology includes: Based on the annotation information of the topology resource pool, the backup links and node resources of the redundant communication network, and the service priority, the evaluation indicators including global robustness, service assurance and reconstruction cost are quantified. The weights of the evaluation indicators are adjusted based on the proportion of redundant links and the proportion of spare resources in the alternative topology, forming an indicator weight matrix. Simulate the random removal of non-highly vulnerable links from alternative topologies, calculate the decrease in connectivity retention rate and service assurance of alternative topologies after removal, and determine whether to add a risk resistance index. The basic evaluation value of the candidate topology is obtained by weighting the index weight matrix, and the comprehensive evaluation value is obtained by superimposing the risk resistance index. The target topology is then selected based on the comprehensive evaluation value.

7. The link self-healing topology control method for redundant communication networks as described in claim 6, characterized in that, The updated output vulnerability index includes: After releasing the link resources of the current topology path, monitor the operation feedback data of the target topology path. The operation feedback data includes path operation parameters and service migration quality. The anti-interference performance of the target topology path's path operation parameters is correlated and compared with the sensitivity of the link to environmental risk scenarios in order to adjust the environmental correction factor. Degradation trends are extracted based on the time-series changes of path operation parameters and similarity matching is performed with historical fault data. The correlation strength coefficient between environmental risk scenarios and performance degradation is adjusted according to the matching results. Based on the service migration quality of different services during parallel transmission, the fusion weights of performance correction features and environmental correction factors are adjusted accordingly. Based on the adjusted environmental correction factor, correlation strength coefficient, and fusion weight, nonlinear fusion is re-executed to dynamically calculate and update the output vulnerability index.

8. The link self-healing topology control method for redundant communication networks as described in claim 7, characterized in that, The parallel transmission includes: Establish parallel transmission channels between the target topology path and the current topology path, allocate transmission resources based on service priority, and initiate data flow migration of the current topology path in stages according to service priority; Real-time monitoring and comparison of the path operation parameters of the two paths, and adjustment of the migration ratio of the data flow based on the degree of degradation of the target topology path relative to the current topology path; Once the data flow of high-priority services has been fully migrated to the target topology path and the data flow of low-priority services has met the preset migration ratio, the link resources of the current topology path will be released in batches according to the connection role of the link.

9. The link self-healing topology control method for redundant communication networks as described in claim 8, characterized in that, Updating the parameters of the weighted evaluation includes: The link load fluctuation and anti-interference performance of the path operation parameters are correlated with global robustness; the remaining link reuse rate of the path operation parameters is correlated with reconstruction costs; and the quality of service migration is correlated with service assurance. The evaluation metrics of the target topology during the selection phase are compared with the corresponding path operation parameters and service migration quality during the parallel transmission phase to obtain the deviation coefficient of the evaluation metrics. The adjustment priority of the indicator weights is determined based on the magnitude of the deviation coefficient, and the indicator weights are adjusted in combination with the deviation coefficient to regenerate the indicator weight matrix for target topology selection.

Images