Network route intelligent switching method and system based on 5G network and physical link

Abstract

The invention relates to the technical field of route switching, in particular to a network route intelligent switching method and system based on a 5G network and a physical link. The method comprises the following steps: deploying routing equipment with double exits to access a physical link network and a 5G network respectively, and collecting link state parameters of the physical link network, wireless transmission parameters of the 5G network and feature parameters of service data flow flowing through the routing equipment in real time through the routing equipment to carry out standardized preprocessing; meanwhile, constructing a network route switching trigger matrix, and calculating a route switching trigger threshold value; starting a millisecond-level route switching mechanism to seamlessly switch to the 5G network when the route switching trigger threshold value is lower than the route switching trigger threshold value; and after the service data flow is switched to the 5G network, continuously monitoring a state recovery parameter corresponding to the physical link network and judging whether a stable operation state is reached, and if so, smoothly switching back to the physical link network. According to the invention, an automatic switching function between a 5G network and a physical link can be realized.

Description

Technical Field

[0001] This invention relates to the field of routing switching technology, and in particular to a method and system for intelligent network routing switching based on 5G network and physical link. Background Technology

[0002] In actual business operations, although the physical lines on which services rely have high transmission quality, they are susceptible to various unforeseen circumstances. Once a physical link is interrupted, the repair time is at least 6 hours, which significantly impacts the user experience. While multiple physical links with different routes can be deployed as backups, this method is costly and is typically used only for backbone line backups, rarely adopted in actual terminal line applications. Currently, 5G networks are widely used as backup lines for customers. However, while 5G networks support flexible deployment and wide-area coverage, they also suffer from signal fluctuations and interference sensitivity. Relying solely on a single type of link is insufficient to meet the continuous transmission requirements of critical services. Summary of the Invention

[0003] Therefore, it is necessary for the present invention to provide a method and system for intelligent switching of network routing based on 5G network and physical link, so as to solve at least one of the above-mentioned technical problems.

[0004] To achieve the above objectives, a method for intelligent network routing switching based on 5G networks and physical links includes the following steps: Step S1: Deploy a routing device with dual exits to access the physical link network and the 5G network respectively. The routing device collects the link status parameters of the physical link network, the wireless transmission parameters of the 5G network, and the service data flow characteristic parameters flowing through the routing device in real time. The collected link status parameters, wireless transmission parameters, and service data flow characteristic parameters are standardized and preprocessed to generate preprocessed physical link status datasets, 5G network transmission datasets, and service characteristic datasets. Step S2: Based on the physical link status dataset, obtain the packet loss rate trend characteristics, latency jitter characteristics, and bandwidth utilization characteristics of the physical link network, and calculate the physical link interruption lag index; based on the 5G network transmission dataset, extract the signal strength stability characteristics, interference level characteristics, and transmission rate fluctuation characteristics of the 5G network, and evaluate and generate the 5G network availability order coefficient; based on the physical link interruption lag index and the 5G network availability order coefficient, construct the network routing handover trigger matrix, and calculate the routing handover trigger threshold; Step S3: Monitor the comparison between the physical link interruption latency index and the route switching trigger threshold in real time. When the physical link interruption latency index is lower than the route switching trigger threshold, start the millisecond-level route switching mechanism to seamlessly switch the service data flow corresponding to the routing device from the physical link network to the 5G network. Step S4: After the service data stream is switched to the 5G network, continuously monitor the status recovery parameters of the physical link network and generate a physical link stability verification report through multiple rounds of bidirectional transmission tests; based on the physical link stability verification report, determine whether the physical link network has reached a stable operating state. If it has, trigger the network route back-switching mechanism to smoothly switch the service data stream corresponding to the routing device from the 5G network back to the physical link network.

[0005] Furthermore, step S1 includes the following steps: Step S11: Establish connections with the physical link network and the 5G network respectively through the dual exit interfaces of the routing device, and configure the data acquisition module to periodically collect the link layer protocol status, frame check error rate and link negotiation status parameters of the physical link network in real time, so as to generate the link status parameters of the physical link network. Step S12: Real-time acquisition of reference signal received power, signal-to-interference-plus-noise ratio, cell handover frequency and network attachment status parameters corresponding to the 5G network through the 5G module to generate wireless transmission parameters of the 5G network. Step S13: Extract the protocol type, packet size distribution, transmission rate, and quality of service requirements parameters of the service data stream flowing through the routing device using deep packet inspection technology, in order to generate the characteristic parameters of the service data stream flowing through the routing device; Step S14: Perform outlier filtering on the collected link status parameters, wireless transmission parameters, and service data stream characteristic parameters to remove pulse-like abnormal data caused by transient interference, and obtain various parameters after outlier filtering. Step S15: Perform time axis alignment processing on various parameters after outlier filtering to unify the data sampling frequency, and transform parameters of different dimensions to the same data range through feature normalization to generate preprocessed physical link status dataset, 5G network transmission dataset and service feature dataset.

[0006] Furthermore, step S2 includes the following steps: Step S21: Extract the data packet transmission records corresponding to the continuous time series from the physical link status dataset, and identify the timestamps and sequence number distributions of lost data packets by comparing the sent data packet sequence and the received data packet sequence; calculate the mean of the packet loss rate sliding window based on the timestamps and sequence number distributions of lost data packets and the clustering density on the time axis, and generate a linear trend coefficient of the packet loss rate as a trend feature of the packet loss rate through trend fitting. Step S22: Extract the transmission delay data of each data packet from the physical link state dataset and calculate the delay difference sequence corresponding to adjacent data packets; generate the delay dispersion parameter by statistically analyzing the standard deviation and coefficient of variation of the delay difference sequence; identify the periodic jitter component by combining the spectrum analysis of the transmission delay data, and use the amplitude ratio between the delay dispersion parameter and the periodic jitter component as the delay jitter feature; Step S23: Calculate the bandwidth occupancy ratio per unit time based on the total bandwidth capacity parameter and real-time transmission rate data corresponding to the physical link network, and extract the peak, valley and average values ​​corresponding to the bandwidth occupancy ratio through a sliding window. At the same time, generate the bandwidth saturation frequency parameter by combining the duration distribution corresponding to the bandwidth occupancy ratio. Use the weighted result of the peak-to-mean ratio, valley-to-mean ratio and bandwidth saturation frequency parameter corresponding to the bandwidth occupancy ratio as the bandwidth utilization feature. Step S24: Calculate the physical link interruption latency index based on the packet loss rate trend characteristics, latency jitter characteristics, and bandwidth utilization characteristics of the physical link network; Step S25: Extract the signal strength stability features, interference level features, and transmission rate fluctuation features of the 5G network based on the 5G network transmission dataset, and evaluate and generate the 5G network availability order coefficient; Step S26: Construct a network routing handover trigger matrix based on the physical link interruption delay index and the 5G network availability order coefficient, and calculate the routing handover trigger threshold.

[0007] Furthermore, step S24 includes the following steps: Step S241: Perform time gradient analysis on the packet loss rate trend characteristics to generate a packet loss deterioration rate parameter by calculating the change in the linear trend coefficient of the packet loss rate per unit time. Combine the critical correlation threshold between packet loss rate and physical link interruption in historical data to map the packet loss deterioration rate parameter to the interruption warning coefficient corresponding to the packet loss dimension. Step S242: Generate a delay jitter frequency based on the delay dispersion parameter and periodic jitter component in the delay jitter characteristics, and calculate the degree of deviation between the delay jitter frequency and the preset service tolerance threshold to generate a delay risk index. At the same time, couple the delay risk index with the retransmission mechanism response speed of the physical link protocol to obtain the interruption warning coefficient corresponding to the delay dimension. Step S243: Analyze the peak-to-mean ratio, valley-to-mean ratio and bandwidth saturation frequency parameters in the bandwidth utilization characteristics to obtain the continuous growth trend corresponding to the bandwidth saturation state, and generate the bandwidth depletion acceleration parameter based on the continuous growth trend corresponding to the bandwidth saturation state. Combine the dynamic adjustment capability parameter corresponding to the physical link bandwidth to obtain the interruption warning coefficient corresponding to the bandwidth dimension. Step S244: The interruption warning coefficients corresponding to packet loss, latency and bandwidth dimensions are fused using the improved DS evidence theory. The weight ratio of each dimension warning coefficient is corrected by introducing a feature conflict resolution factor to generate a physical link comprehensive degradation index. Step S245: Based on the physical link comprehensive degradation index and combined with the time distribution of physical links from the same degree of degradation to complete interruption in historical data, the interruption delay assessment calculation is performed to obtain the physical link interruption delay index.

[0008] Furthermore, step S245 includes the following steps: By extracting all complete link degradation-disruption event sequences from the historical operation logs of physical links, each sequence contains feature parameters of the entire process from the starting point of the same degree of degradation to complete interruption, and classifying these sequences according to the evolution curve shape of the physical link comprehensive degradation index, several typical degradation modes are generated, with each mode corresponding to a set of feature parameter distribution patterns. The time series of the comprehensive degradation index corresponding to the current physical link is matched with various typical degradation modes to calculate the dynamic time warping distance between the current sequence and each typical degradation mode. The typical degradation mode with the smallest distance is selected as the benchmark reference mode. At the same time, the subsequent time evolution parameters of the nodes with the same degradation degree as the current nodes in the benchmark reference mode are extracted. Based on the historical data statistics corresponding to the benchmark reference mode, the time sample set from nodes with the same degree of degradation to complete interruption is statistically analyzed, and the time distribution probability density function is generated by kernel density estimation. At the same time, the expected time value and confidence interval parameter corresponding to the function are calculated to obtain the historical average time delay benchmark value. By introducing a real-time environmental correction factor corresponding to the current physical link, which is calculated based on the temperature drift parameter, electromagnetic interference intensity and link load fluctuation characteristics of the current physical link, and converting the real-time environmental correction factor into a time adjustment coefficient through nonlinear mapping, the historical average time delay benchmark value is dynamically corrected based on the time adjustment coefficient to generate the corrected time delay benchmark value. The instantaneous rate of change of the current comprehensive degradation index is generated by using the subsequent time evolution parameters of nodes with the same degree of current degradation. The acceleration parameters corresponding to the link degradation process are calculated by combining the corrected time delay benchmark value. At the same time, the acceleration parameters are coupled with the corrected time delay benchmark value to obtain the predicted remaining time of the physical link from the current state to complete interruption. After normalization, the physical link interruption time delay index is generated.

[0009] Furthermore, step S25 includes the following steps: Extract continuous time series reference signal received power data from 5G network transmission datasets, calculate the mean drift and variance rate of change of signal strength, combine signal quality mutation point detection to identify step change events of signal strength, and weight and fuse the mean drift, variance rate of change and the occurrence frequency of step change events to generate signal strength stability features. Based on the signal-to-interference-plus-noise ratio in the 5G network transmission data set, the spectral distribution characteristics of the interference signal are separated, and the instantaneous frequency and energy ratio of the interference signal are extracted by wavelet transform. The distribution entropy value of the interference signal in different frequency bands is calculated. Combined with the interference duration and the overlap parameter of the service frequency band, the interference level assessment matrix is ​​constructed, and the eigenvalues ​​of the interference level assessment matrix are used as the interference level features. Real-time sampling sequences of transmission rate are extracted from 5G network transmission datasets. Rate fluctuation sequences are generated by calculating the rate change of adjacent sampling points. The rate fluctuation sequences are decomposed into several intrinsic mode components. The periodic characteristics and energy proportion of each component are analyzed. The dominant fluctuation component is selected and its amplitude variation coefficient is calculated. The amplitude variation coefficient is coupled with the Hurst exponent of the rate fluctuation to generate transmission rate fluctuation characteristics. A three-dimensional space for 5G network availability assessment is constructed, with signal strength stability characteristics, interference level characteristics, and transmission rate fluctuation characteristics as coordinate axes. Availability benchmark points in the space are calibrated using historical communication quality data. The Mahalanobis distance between the current network status and the benchmark point in the three-dimensional space is calculated, and this distance is converted into an initial availability score. A network load resilience coefficient is introduced to correct the initial availability score. This coefficient is calculated based on the current number of user connections, cell handover success rate and resource block utilization of the 5G network. By establishing a mapping relationship between the network load resilience coefficient and the availability decay factor, the initial availability score is dynamically adjusted to generate the 5G network availability order coefficient.

[0010] Furthermore, step S26 includes the following steps: Extract the exponential change slope parameter from the time series of the physical link interruption delay index, and generate the link deterioration urgency coefficient by analyzing the absolute value growth trend of the exponential change slope parameter. At the same time, based on the historical fluctuation range of the 5G network availability order coefficient, calculate the deviation parameter between the current availability order coefficient and the optimal value to obtain the 5G network service capacity margin. A two-dimensional feature mapping space is constructed, with the link deterioration urgency coefficient as the vertical axis and the 5G network service capacity margin as the horizontal axis. The historical handover decision data is mapped into several decision-dense regions in the space through the K-means clustering algorithm. Each dense region corresponds to a set of typical handover strategy labels, generating a decision reference map. Based on the boundary parameters of each dense area in the decision reference map, a network routing handover trigger matrix is ​​established. The row dimension of the matrix corresponds to the graded interval of the urgency of link deterioration, the column dimension corresponds to the graded interval of the 5G network service capacity margin, and the matrix element is the coupling result of the historical handover success rate and the service interruption risk value under the combination of the interval. A transmission demand sensitivity parameter corresponding to the business characteristic dataset is introduced. This parameter is calculated based on the real-time weight of the business, data integrity requirements and bandwidth dependence. The transmission demand sensitivity parameter is converted into a matrix correction factor through fuzzy hierarchical analysis to dynamically adjust the element values ​​in the network routing switching trigger matrix. A multi-objective optimization algorithm is used to solve the Pareto optimal solution of the adjusted matrix. The weighted product of the link deterioration urgency coefficient and the 5G network service capacity margin is used as the objective function. Combined with the constraint of service interruption risk, the critical decision thresholds under different business scenarios are calculated. These critical decision thresholds are normalized and fused to generate the routing handover trigger threshold.

[0011] Furthermore, step S4 includes the following steps: Step S41: After the service data stream is switched to the 5G network, continuously monitor and collect the status recovery parameters corresponding to the physical link network, including link connectivity parameters, data forwarding success rate and protocol negotiation status parameters; Step S42: Determine whether the physical link network has recovered from the fault state based on the collected state recovery parameters. When the link connectivity parameters meet the preset conditions and the data forwarding success rate reaches the recovery threshold, it is determined that the physical link network has entered the state recovery period. Step S43: During the state recovery period, start the link stability verification process to conduct multiple rounds of bidirectional transmission tests by sending test data packets of different sizes and protocol types, and record the packet loss rate, latency and jitter parameters of each round of test; Step S44: Calculate the stability score of the physical link network based on the packet loss rate, latency and jitter parameters of multiple rounds of testing. When the stability score of multiple consecutive rounds of testing is higher than the stability threshold, generate a physical link stability verification report. The physical link stability verification report includes the link recovery time, stability and recommended back-off timing parameters. Step S45: Based on the physical link stability verification report, determine whether the physical link network has reached a stable operating state. If it has, trigger the network route back-switching mechanism to smoothly switch the service data flow corresponding to the routing device from the 5G network back to the physical link network.

[0012] Furthermore, step S45 includes the following steps: The system analyzes the parameters in the physical link stability verification report. When the stability of the physical link network reaches the preset back-off standard and the recommended back-off timing parameter indicates that the current time is the best back-off window, the back-off preparation process is started. In the back-switching preparation process, a data transmission channel corresponding to the physical link network is pre-established, and the service data streams on the 5G network are sorted based on the service priority parameters in the service feature dataset, with priority given to preparing back-switching for real-time interactive services. The service priority is corrected based on the physical link interruption time delay index as the time delay impact factor. A gradual back-switch strategy is adopted to switch the service data stream from the 5G network back to the physical link network in order of the corrected service priority from high to low. After each batch of service switching is completed, the continuity and integrity parameters of service transmission are detected. Once all service data streams have successfully switched back to the physical link network and the transmission status is stable, the data transmission channel of the 5G network is shut down. The total duration of this switchback process, the number of service interruptions, and the amount of data loss are recorded. These parameters are then merged with the switchback process log to optimize the threshold parameters and weight coefficients in the millisecond-level routing switchback mechanism.

[0013] Furthermore, the present invention also provides a network routing intelligent handover system based on 5G network and physical link, used to execute the network routing intelligent handover method based on 5G network and physical link as described above. The network routing intelligent handover system based on 5G network and physical link includes: The routing device data processing module is used to deploy routing devices with dual exits to access the physical link network and the 5G network respectively. It collects link status parameters of the physical link network, wireless transmission parameters of the 5G network, and service data flow characteristic parameters flowing through the routing device in real time through the routing device. It performs standardized preprocessing on the collected link status parameters, wireless transmission parameters, and service data flow characteristic parameters to generate preprocessed physical link status datasets, 5G network transmission datasets, and service characteristic datasets. The routing handover trigger evaluation module is used to obtain the packet loss rate trend characteristics, latency jitter characteristics, and bandwidth utilization characteristics of the physical link network based on the physical link status dataset, and calculate the physical link interruption lag index; extract the signal strength stability characteristics, interference level characteristics, and transmission rate fluctuation characteristics of the 5G network based on the 5G network transmission dataset, and evaluate and generate the 5G network availability order coefficient; construct the network routing handover trigger matrix based on the physical link interruption lag index and the 5G network availability order coefficient, and calculate the routing handover trigger threshold. The network routing link switching module is used to monitor the comparison between the physical link interruption latency index and the routing switching trigger threshold in real time. When the physical link interruption latency index is lower than the routing switching trigger threshold, a millisecond-level routing switching mechanism is activated to seamlessly switch the service data flow corresponding to the routing device from the physical link network to the 5G network. The network routing back-off response module is used to continuously monitor the status recovery parameters of the physical link network after the service data flow is switched to the 5G network and generate a physical link stability verification report through multiple rounds of bidirectional transmission tests. Based on the physical link stability verification report, it determines whether the physical link network has reached a stable operating state. If it has, it triggers the network routing back-off mechanism to smoothly switch the service data flow corresponding to the routing device from the 5G network back to the physical link network.

[0014] The beneficial effects of this invention are: 1. The intelligent network routing switching method based on 5G network and physical link proposed in this invention, compared with the prior art, has the following advantages: by deploying dual-exit routing equipment to access both the physical link and the 5G network, and by standardizing and preprocessing the collected multi-type parameters, it solves the problems of insufficient data dimensions and chaotic parameter formats caused by single-link data collection, which lead to difficulties in subsequent analysis. The dual-exit design constructs the data collection foundation of "primary and backup links" from the source, ensuring that the real-time status of both types of links can be obtained simultaneously, avoiding the limitations of single-link monitoring; while the standardized preprocessing eliminates the differences in dimensions and format conflicts of different types of parameters (such as link status, wireless transmission, and service data flow), generating a structured dataset. This avoids untimely switching due to data lag and effectively compensates for the shortcomings of traditional single-link monitoring in predicting faults in advance. Secondly, by extracting the core features of the two types of links, calculating key indices and coefficients, and constructing a switching trigger matrix and threshold, it solves the problem that traditional backup link switching relies on manual judgment and lacks quantitative standards, resulting in inaccurate switching timing. By extracting packet loss rate trends and latency jitter characteristics from physical links and calculating the interruption lag index, stability can be accurately quantified. For 5G networks, features such as signal stability and interference levels are extracted, and availability order coefficients are generated, allowing for an objective assessment of transmission capabilities. Based on this, a trigger matrix and thresholds transform abstract link states into quantifiable switching criteria, avoiding simplistic judgments relying solely on physical link interruptions or 5G signal strength. Compared to the traditional passive "switching only after network outage" mode, this step achieves a proactive assessment of link states, providing scientific decision support for subsequent millisecond-level switching and effectively reducing the risk of service interruption due to improper switching timing. Then, by monitoring the comparison results of the index and thresholds in real time, a millisecond-level switching mechanism is initiated to switch service flows from the physical link to the 5G network, solving the problems of long recovery times (minimum 6 hours) after physical link interruptions and long-term service interruptions caused by slow traditional switching responses. Real-time comparison ensures that critical abnormal states of physical links (interruption latency index below a threshold) can be captured immediately, avoiding waiting until a complete interruption to trigger a switchover. Millisecond-level switching speed significantly reduces the interruption time of service flows during link switching, far lower than the switching time of manual intervention or traditional equipment. Simultaneously, the seamless switching design ensures the continuity of service data transmission, avoiding packet loss or out-of-order delivery during switching, effectively mitigating the impact of physical link interruptions on user experience. Finally, by continuously monitoring physical link recovery parameters, generating stability verification reports, and triggering a back-switch mechanism, the problem of "long-term occupation of the backup link and inability to switch back to the high-quality primary link in a timely manner" after traditional backup link switching is solved. Continuous monitoring of the physical link status ensures timely detection of recovery signs, avoiding the long-term service burden on the 5G backup link due to neglecting primary link recovery—avoiding transmission risks caused by 5G signal fluctuations and interference sensitivity, while fully utilizing the high transmission quality of the physical link.Verification reports generated from multiple rounds of bidirectional transmission tests can scientifically verify whether the physical link has truly reached a stable operating state, avoiding repeated switching caused by short-term recovery and reconnection. The smooth reconnection mechanism ensures the continuity of transmission when the service flow switches back from the 5G network to the physical link, preventing data interruption or lag. This closed-loop design realizes the dynamic optimization scheduling of the "primary and backup links", which not only solves the emergency transmission problem after the physical link is interrupted, but also avoids the drawbacks of long-term use of backup links, further ensuring the stability and efficiency of critical service transmission.

[0015] 2. The intelligent network routing handover system based on 5G network and physical link proposed in this invention is composed of a routing device data processing module, a routing handover trigger evaluation module, a network routing link handover module, and a network routing back-off response module. It can realize any intelligent network routing handover method based on 5G network and physical link as described in this invention. It is used to combine the operations between the computer programs running on each module to realize the intelligent network routing handover method based on 5G network and physical link. The internal structure of the system cooperates with each other, which can greatly reduce repetitive work and manpower input, and can quickly and effectively provide a more accurate and efficient intelligent network routing handover process based on 5G network and physical link, thereby simplifying the operation process of the intelligent network routing handover system based on 5G network and physical link. Attached Figure Description

[0016] Other features, objects, and advantages of the invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings: Figure 1 This is a flowchart illustrating the steps of the intelligent network routing switching system based on 5G network and physical link of the present invention. Figure 2 for Figure 1 A detailed flowchart of step S1. Detailed Implementation

[0017] The technical system of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0018] To achieve the above objectives, please refer to Figures 1 to 2 This invention provides a method for intelligent network routing switching based on 5G networks and physical links. For embodiments of this invention, please refer to... Figure 1The diagram illustrates the steps of intelligent network routing switching based on 5G network and physical link according to the present invention. In this example, the intelligent network routing switching method based on 5G network and physical link includes the following steps: Step S1: Deploy a routing device with dual exits to access the physical link network and the 5G network respectively. The routing device collects the link status parameters of the physical link network, the wireless transmission parameters of the 5G network, and the service data flow characteristic parameters flowing through the routing device in real time. The collected link status parameters, wireless transmission parameters, and service data flow characteristic parameters are standardized and preprocessed to generate preprocessed physical link status datasets, 5G network transmission datasets, and service characteristic datasets. In this embodiment of the invention, a routing device with dual gigabit Ethernet exits (GE0 / 0 and GE0 / 1) is deployed. The GE0 / 0 interface is connected to a gigabit switch in the physical link network (the physical link is a fiber optic backbone network with a total bandwidth of 1Gbps) via Category 6 unshielded twisted-pair cable. The GE0 / 1 interface is connected to a 5G base station via a USB-to-5G module (supporting SA networking, frequency band n41). The routing device's data acquisition module is configured with a acquisition frequency of 10 seconds / time: physical link status parameters include frame check error rate, link negotiation status, and real-time transmission rate; 5G network wireless transmission parameters include reference signal received power (RSRP), signal-to-interference-plus-noise ratio (SINR), and cell handover frequency; and service data stream characteristic parameters include protocol type (HTTP / HTTPS / FTP), packet size distribution, and quality of service (QoS) level. The collected data underwent standardized preprocessing: Impulse outliers in the frame check error rate (e.g., 0.5% exceeding the mean of 0.03% + 3 × standard deviation of 0.02%) were removed using the 3σ criterion; RSRP (-120 to -70dBm) was converted to the 0-1 range (-85dBm corresponds to 0.7) using Min-Max normalization; the sampling frequency of each parameter was standardized to 10 seconds / time (5G parameters were originally 5 seconds / time, but downsampling was used), ultimately generating three preprocessed datasets, each stored in the format "timestamp-parameter1-parameter2-…-parametern".

[0019] Step S2: Based on the physical link status dataset, obtain the packet loss rate trend characteristics, latency jitter characteristics, and bandwidth utilization characteristics of the physical link network, and calculate the physical link interruption lag index; based on the 5G network transmission dataset, extract the signal strength stability characteristics, interference level characteristics, and transmission rate fluctuation characteristics of the 5G network, and evaluate and generate the 5G network availability order coefficient; based on the physical link interruption lag index and the 5G network availability order coefficient, construct the network routing handover trigger matrix, and calculate the routing handover trigger threshold; In this embodiment of the invention, data from the physical link status dataset for nearly one hour is extracted to calculate the packet loss rate trend characteristics: using a 50-second sliding window, the packet loss rate is 0.01% from the 1st to the 50th second, 0.012% from the 11th to the 60th second, and 0.03% from the 511th to the 560th second. The linear trend coefficient is 0.00004% / second obtained by fitting with the least squares method. The latency jitter characteristics are calculated as follows: the standard deviation of the latency difference between adjacent data packets is 0.7ms, the coefficient of variation is 3.5, and the 0.02Hz periodic jitter component (amplitude 2ms) identified by spectrum analysis is combined to obtain an amplitude ratio of 0.35. The bandwidth utilization characteristics are as follows: peak value 90%, mean 82%, valley value 78%, bandwidth saturation frequency 0.05 times / minute, weighted by weights of 0.4 / 0.3 / 0.3 to obtain 0.739. The physical link interruption latency index is calculated using the formula "Physical link interruption latency index = 1 - (0.00004 × 10000 × 0.3 + 0.35 × 0.3 + (1 - 0.739) × 0.4)", resulting in 0.701. Signal strength stability feature (0.715), interference level feature (0.65), and transmission rate fluctuation feature (0.72) are extracted from the 5G network transmission dataset. These are weighted with weights of 0.4 / 0.3 / 0.3 to obtain an availability order coefficient of 0.697. A 3×3 routing handover trigger matrix is ​​constructed (rows: latency index 0-0.3 / 0.3-0.7 / 0.7-1.0; columns: availability 0-0.4 / 0.4-0.8 / 0.8-1.0). Matrix elements (e.g., the element corresponding to latency 0.7-1.0 and availability 0.4-0.8 is 0.772) are calculated. The NSGA-II algorithm is used to solve for the Pareto optimal solution, and the trigger threshold is obtained after normalization and fusion, resulting in 0.87.

[0020] Step S3: Monitor the comparison between the physical link interruption latency index and the route switching trigger threshold in real time. When the physical link interruption latency index is lower than the route switching trigger threshold, start the millisecond-level route switching mechanism to seamlessly switch the service data flow corresponding to the routing device from the physical link network to the 5G network. In this embodiment of the invention, the physical link interruption latency index is obtained every 10 seconds by the real-time monitoring module of the routing device and compared with the trigger threshold of 0.87: the index is 0.701 < 0.87 for the first time (10 seconds), 0.72 < 0.87 for the second time (20 seconds), and 0.73 < 0.87 for the third time (30 seconds). If the index is lower than the threshold for three consecutive times, the millisecond-level routing switching mechanism is activated. The routing device switches the service data flow from the physical link GE0 / 0 interface (IP192.168.1.1) to the 5G network GE0 / 1 interface (IP192.168.2.1) through a hardware fast forwarding channel: First, it buffers the currently incomplete industrial control service data packets (about 100KB), modifies the routing table entry (switching the default gateway from 192.168.1.254 to 192.168.2.254), and then prioritizes forwarding industrial control services according to QoS level (EF level). The switching process takes 8ms, with no service interruption or data loss, completing a seamless switch.

[0021] Step S4: After the service data stream is switched to the 5G network, continuously monitor the status recovery parameters of the physical link network and generate a physical link stability verification report through multiple rounds of bidirectional transmission tests; based on the physical link stability verification report, determine whether the physical link network has reached a stable operating state. If it has, trigger the network route back-switching mechanism to smoothly switch the service data stream corresponding to the routing device from the 5G network back to the physical link network.

[0022] In this embodiment of the invention, after switching services to the 5G network, the routing device collects physical link status recovery parameters every minute: link connectivity (ICMP response success rate), data forwarding success rate, and protocol negotiation status. At the 5th minute, the ICMP response success rate is 100%, the forwarding success rate is 99.5%, and the negotiation status is stable at 1000Mbps full-duplex, indicating the start of the status recovery period. Five rounds of bidirectional transmission testing are initiated (5 minutes per round): Round 1 sends 64-byte TCP packets with a packet loss rate of 0.05%, latency of 12ms, and jitter of 2ms; Rounds 2-5 send 1500-byte TCP packets, 64-byte UDP packets, 1500-byte UDP packets, and mixed packets respectively. The stability score exceeds 90 points for four consecutive rounds, generating a stability verification report (recovery time at the 5th minute, stability 93.4%, recommended traffic percentage at the time of switchback 30%). The analysis report determined that the physical link was stable and triggered the rollback mechanism: the service was switched gradually according to the service priority (industrial control → video → file), with 20% of the traffic switched in each batch. After the switch was completed, the transmission status was checked (latency 13-15ms, packet loss rate 0%), the 5G channel was closed, and the smooth rollback was completed.

[0023] Furthermore, as an embodiment of the present invention, reference is made to... Figure 2 As shown, Figure 1A detailed flowchart of step S1 is shown below. In this embodiment, step S1 includes the following steps: Step S11: Establish connections with the physical link network and the 5G network respectively through the dual exit interfaces of the routing device, and configure the data acquisition module to periodically collect the link layer protocol status, frame check error rate and link negotiation status parameters of the physical link network in real time, so as to generate the link status parameters of the physical link network. In this embodiment of the invention, a routing device with dual gigabit Ethernet ports is selected, wherein the GE0 / 0 interface is connected to a gigabit switch in the physical link network (the switch is connected to the enterprise core router, and the physical link is a fiber optic backbone network) via a Category 6 unshielded twisted pair cable, and the GE0 / 1 interface is connected to the 5G base station signal via a USB adapter 5G module (supporting SA standalone networking mode), thereby completing the establishment of dual network connections. The data acquisition module (integrated in the router device firmware) is configured to acquire data every 10 seconds, collecting three parameters of the physical link network: the link layer protocol status is obtained by parsing the LLDP protocol messages of the GE0 / 0 interface (normally displaying "Protocol UP, negotiation rate 1000Mbps full duplex", abnormally displaying "Protocol DOWN or rate negotiation failed"); the frame check error rate is calculated by counting the number of CRC error frames received by the GE0 / 0 interface within 10 seconds (e.g., 10000 total frames in 10 seconds, 3 error frames), and is 0.03% according to the formula "error rate = number of error frames / total number of frames × 100%"; the link negotiation status parameter is obtained by reading the interface register to obtain the negotiation priority ("1000Mbps full duplex > 100Mbps full duplex > 10Mbps half duplex") and the current negotiation result. After each collection, the three parameters are encapsulated in the format of "timestamp-protocol status-error rate-negotiation result" to generate the link status parameters of the physical link network (such as "20240615143000-UP-0.03%-1000Mbps full duplex").

[0024] Step S12: Real-time acquisition of reference signal received power, signal-to-interference-plus-noise ratio, cell handover frequency and network attachment status parameters corresponding to the 5G network through the 5G module to generate wireless transmission parameters of the 5G network. In this embodiment of the invention, the 5G module of the routing device accesses the operator's 5G network (frequency band n41, bandwidth 100MHz) via a Nano-SIM card, and the parameter acquisition frequency of the 5G module is configured to be 5 seconds / time. The Reference Received Power (RSRP) is measured by the module's built-in RF chip to measure the downlink reference signal strength of the base station, with an output value ranging from -120dBm to -70dBm (e.g., a collected value of -85dBm indicates a strong signal). The Signal-to-Interference-plus-Noise Ratio (SINR) is calculated by the module's signal processing unit to determine the ratio of useful signal power to interference-plus-noise power, with an output value ranging from -20dB to 30dB (e.g., a collected value of 15dB indicates good signal quality). The cell handover frequency is determined by counting the number of times the module's cell PCI (Physical Cell Identifier) ​​changes within one hour (e.g., two handovers within one hour indicate stable network coverage). Network attachment status parameters are obtained by querying the module's NAS layer signaling (normally displaying "EPS attachment successful, registered state EMM-REGISTERED", abnormally displaying "attachment failed or registered state EMM-DEREGISTERED"). After each acquisition, the four parameters are recorded in the format of "timestamp-RSRP-SINR-switching frequency-attachment status" to generate the wireless transmission parameters of the 5G network (such as "20240615143000--85dBm-15dB-2 times / hour-attachment successful").

[0025] Step S13: Extract the protocol type, packet size distribution, transmission rate, and quality of service requirements parameters of the service data stream flowing through the routing device using deep packet inspection technology, in order to generate the characteristic parameters of the service data stream flowing through the routing device; In this embodiment of the invention, a deep packet inspection (DPI) engine is deployed in the traffic processing unit of the routing device. The engine identifies the characteristics of the service data flow by parsing the IP packet headers and application layer payloads flowing through the device. Protocol type identification is performed by matching the port number of the data packets (e.g., ports 80 / 443 correspond to HTTP / HTTPS protocols, and port 21 corresponds to FTP protocols) and application layer signatures (e.g., the "GET / POST" request header of HTTP data packets) to determine that a data stream is an "HTTPS protocol". Data packet size distribution is generated by statistically analyzing the data packet lengths of the data stream within one minute (e.g., out of 100 data packets, 10 are under 64 bytes, 30 are 64-127 bytes, 40 are 128-255 bytes, and 20 are over 256 bytes) to generate a size distribution ratio. Transmission rate is calculated by calculating the total number of bytes in the data stream within one minute (e.g., 10MB is 1.39Mbps). Quality of Service (QoS) requirements are determined by parsing the DSCP markers of the data packets (e.g., a DSCP value of EF corresponds to voice services requiring low latency; a DSCP value of AF41 corresponds to video services requiring high bandwidth) to determine the QoS level of the HTTPS data stream as "Guaranteed Bandwidth". The above operations are repeated for each type of service data stream to generate service data stream characteristic parameters containing "protocol type - size distribution - rate - QoS requirements".

[0026] Step S14: Perform outlier filtering on the collected link status parameters, wireless transmission parameters, and service data stream characteristic parameters to remove pulse-like abnormal data caused by transient interference, and obtain various parameters after outlier filtering. In this embodiment of the invention, an outlier filtering tool based on the 3σ criterion is used to process three types of parameters. For the frame check error rate (normal range 0-0.1%) in the link status parameters, the mean of 100 historical data collections is calculated to be 0.03%, the standard deviation to be 0.01%, and the upper limit of 3σ is 0.06%. If a collection value is 0.12% (exceeding the upper limit), it is determined to be pulse-type abnormal data (caused by transient electromagnetic interference) and is discarded. For the RSRP (normal range -110 to -75 dBm) in the wireless transmission parameters, the mean is -85 dBm, the standard deviation to be 5 dB, and the lower limit of 3σ is -100 dBm. If a collection value is -125 dBm (exceeding the lower limit), it is determined to be abnormal and discarded. For the transmission rate (normal range 1-2 Mbps for a certain HTTPS stream) in the service data stream characteristic parameters, the mean is 1.5 Mbps, the standard deviation to be 0.2 Mbps, and the upper limit of 3σ is 2.1 Mbps. If a collection value is 3.5 Mbps (exceeding the upper limit), it is determined to be abnormal and discarded. After removing outliers, linear interpolation is used to fill in the missing data (e.g., if the data at 14:30:00 is outlier, interpolate the 1.4Mbps at 14:29:50 and the 1.6Mbps at 14:30:10 to the 1.5Mbps at 14:30:00), and obtain various parameters after outlier filtering.

[0027] Step S15: Perform time axis alignment processing on various parameters after outlier filtering to unify the data sampling frequency, and transform parameters of different dimensions to the same data range through feature normalization to generate preprocessed physical link status dataset, 5G network transmission dataset and service feature dataset.

[0028] In this embodiment of the invention, by using a time axis alignment tool, with a uniform sampling frequency of 5 seconds (5G parameters have a raw frequency of 5 seconds, link parameters need to be interpolated at 10 seconds, and service parameters need to be downsampled at 1 minute): the 10-second data of link status parameters (14:30:00, 14:30:10) are inserted with the average of the two at 14:30:05; the 1-minute data of service parameters (14:30:00) is copied into 12 identical data at 5-second intervals. After alignment, the Min-Max normalization method is used to convert different dimensional parameters to the [0,1] range: the frame check error rate (0-0.1%) is calculated according to "normalized value = (actual value - 0) / (0.1 - 0)", and 0.03% corresponds to 0.3; RSRP (-120 to -70dBm) is calculated according to "normalized value = (actual value - (-120)) / (-70 - (-120))", and -85dBm corresponds to 0.7; the transmission rate (1-2Mbps) is calculated according to "normalized value = (actual value - 1) / (2 - 1)", and 1.5Mbps corresponds to 0.5. Finally, preprocessed physical link state datasets (including normalized protocol status, error rate, and negotiation status, 5 seconds / database), 5G network transmission datasets (including normalized RSRP, SINR, etc., 5 seconds / database), and service feature datasets (including normalized rate, QoS requirements, etc., 5 seconds / database) are generated for subsequent route switching decisions.

[0029] Furthermore, step S2 includes the following steps: Step S21: Extract the data packet transmission records corresponding to the continuous time series from the physical link status dataset, and identify the timestamps and sequence number distributions of lost data packets by comparing the sent data packet sequence and the received data packet sequence; calculate the mean of the packet loss rate sliding window based on the timestamps and sequence number distributions of lost data packets and the clustering density on the time axis, and generate a linear trend coefficient of the packet loss rate as a trend feature of the packet loss rate through trend fitting. In this embodiment of the invention, data packet transmission records for one hour are extracted from the physical link status dataset (1Gbps fiber optic link, sampling frequency 1 time / second). The sequence numbers of the sent data packets are 1-3600 (1 data packet per second), and the sequence numbers of the received data packets are 1-50, 52-200, and 202-3600. The lost data packet sequence numbers 51 and 201 are identified by sequence number comparison, and the corresponding timestamps are the 51st second and the 201st second. The cluster density of lost data packets on the time axis was calculated (two packet losses were scattered over 3600 seconds, with a density of 0.00056 packets / second). A 50-second sliding window was used to calculate the packet loss rate: no packet loss in the first 1-50 second window, packet loss rate 0%; one packet loss in the 51-100 second window, packet loss rate 2%; no packet loss in the 101-150 second window, packet loss rate 0%; no packet loss in the 151-200 second window, packet loss rate 0%; and one packet loss in the 201-250 second window, packet loss rate 2%. A sliding window mean sequence was generated. The mean sequence was linearly fitted using the least squares method. The slope of the fitted line was 0.00004% / second. This slope is the linear trend coefficient of the packet loss rate, which, together with the sliding window mean, constitutes the trend feature of the packet loss rate.

[0030] Step S22: Extract the transmission delay data of each data packet from the physical link state dataset and calculate the delay difference sequence corresponding to adjacent data packets; generate the delay dispersion parameter by statistically analyzing the standard deviation and coefficient of variation of the delay difference sequence; identify the periodic jitter component by combining the spectrum analysis of the transmission delay data, and use the amplitude ratio between the delay dispersion parameter and the periodic jitter component as the delay jitter feature; In this embodiment of the invention, the transmission delay data (unit: ms) of 3600 data packets within one hour is extracted from the physical link state dataset, such as the delay of the 1st data packet being 12, the 2nd being 13, the 3rd being 11, the 4th being 14… and the 3600th being 15. The delay difference sequence of adjacent data packets is calculated: 13-12=1, 11-13=-2, 14-11=3…15-the 3599th delay value. The standard deviation of this difference sequence is calculated: the mean of all differences is 0.2ms, the sum of the squares of the deviations of each difference from the mean is 1800, and the standard deviation = √(1800 / 3599) ≈ 0.707ms; the coefficient of variation = standard deviation / mean = 0.707 / 0.2 ≈ 3.535. Both together constitute the delay dispersion parameter. Fourier transform spectrum analysis was performed on the transmission delay data to identify a periodic jitter component with a frequency of 0.02 Hz (period of 50 seconds) and an amplitude of 2 ms. The ratio of the delay dispersion parameter (standard deviation 0.707 ms) to the amplitude of the periodic jitter component (2 ms) was calculated to be 0.707 / 2 = 0.3535. This ratio represents the delay jitter characteristic.

[0031] Step S23: Calculate the bandwidth occupancy ratio per unit time based on the total bandwidth capacity parameter and real-time transmission rate data corresponding to the physical link network, and extract the peak, valley and average values ​​corresponding to the bandwidth occupancy ratio through a sliding window. At the same time, generate the bandwidth saturation frequency parameter by combining the duration distribution corresponding to the bandwidth occupancy ratio. Use the weighted result of the peak-to-mean ratio, valley-to-mean ratio and bandwidth saturation frequency parameter corresponding to the bandwidth occupancy ratio as the bandwidth utilization feature. In this embodiment of the invention, the total bandwidth capacity parameter of the physical link network is 1Gbps (1000Mbps). Real-time transmission rate data (unit: Mbps) is extracted within one hour, such as 800 for the first minute, 850 for the second minute, 780 for the third minute, 900 for the fourth minute, and so on, up to 920 for the 60th minute. The bandwidth occupancy ratio per unit time is calculated as follows: 800 / 1000=80%, 850 / 1000=85%, 780 / 1000=78%, 900 / 1000=90%, and so on, up to 920 / 1000=92%. A 10-minute sliding window is used to extract features: within the first 10 minutes, the peak is 90% (4th minute), the valley is 78% (3rd minute), and the average level is 82%; within the 11th to 20th minute window, the peak is 92%, the valley is 80%, and the average level is 85%, and so on. The duration distribution of bandwidth occupancy ≥90% was analyzed: it occurred 3 times within 1 hour, with durations of 5 minutes, 8 minutes, and 6 minutes respectively. The bandwidth saturation frequency parameter = 3 times / 60 minutes = 0.05 times / minute. The peak-to-mean ratio was calculated as 90% / 82% ≈ 1.097, and the valley-to-mean ratio as 78% / 82% ≈ 0.951. Using the weighted formula "bandwidth utilization characteristic = 1.097 × 0.4 + 0.951 × 0.3 + 0.05 × 0.3", the result is 0.4388 + 0.2853 + 0.015 = 0.7391.

[0032] Step S24: Calculate the physical link interruption latency index based on the packet loss rate trend characteristics, latency jitter characteristics, and bandwidth utilization characteristics of the physical link network; In this embodiment of the invention, the linear trend coefficient in the packet loss rate trend feature is weighted at 0.3, the sliding window mean (0.055% over 1 hour) at 0.2, the latency jitter feature at 0.3, and the bandwidth utilization feature at 0.2. The physical link interruption latency index is calculated using the formula: "Physical link interruption latency index = 1 - ((linear trend coefficient × 10000 × 0.3) + (sliding window mean × 2 × 0.2) + (latency jitter feature × 0.3) + (1 - bandwidth utilization feature) × 0.2)". The calculated values ​​are: linear trend coefficient 0.00004% / second × 10000 = 0.4, sliding window mean 0.055% × 2 = 0.11, latency jitter feature 0. 0.3535, 1 - bandwidth utilization characteristic 0.2609, substituting, we get 1 - (0.4×0.3+0.11×0.2+0.3535×0.3+0.2609×0.2) = 1 - (0.12+0.022+0.10605+0.05218) = 1 - 0.29923 = 0.70077, which is the physical link interruption latency index (range 0-1).

[0033] Step S25: Extract the signal strength stability features, interference level features, and transmission rate fluctuation features of the 5G network based on the 5G network transmission dataset, and evaluate and generate the 5G network availability order coefficient; In this embodiment of the invention, one hour of signal strength data (RSRP, unit: dBm) is extracted from the 5G network transmission dataset (frequency band n41, sampling frequency 1 time / minute): -85, -86, -84, -87…-88. The mean drift is calculated as -88 - (-85.5) = -2.5 dBm, the variance change rate is (3.5 - 2.8) / 2.8 × 100% ≈ 25%, the step change event frequency is 0.03 times / minute, and the signal strength stability feature is generated according to the formula: 1 - (2.5 / 10 × 0.4 + 25 / 100 × 0.3 + 0.03 / 0.1 × 0.3) = 0.715. The interference signal spectrum is extracted, the distribution entropy value is calculated as 0.45, the interference duration is 35 minutes / hour, the overlap is 25%, and the interference level feature is generated as 0.65. The transmission rate sequence was extracted, and the amplitude variation coefficient of the dominant fluctuation component was calculated to be 0.22, the Hurst exponent to be 0.68, and the transmission rate fluctuation characteristic to be 0.72 was generated. Using the weighted formula “5G network availability order coefficient = 0.715 × 0.4 + 0.65 × 0.3 + 0.72 × 0.3”, the result was 0.286 + 0.195 + 0.216 = 0.697.

[0034] Step S26: Construct a network routing handover trigger matrix based on the physical link interruption delay index and the 5G network availability order coefficient, and calculate the routing handover trigger threshold.

[0035] In this embodiment of the invention, a 3×3 network routing switching trigger matrix is ​​constructed by dividing the physical link interruption latency index into three intervals: low (0-0.3), medium (0.3-0.7), and high (0.7-1.0); and the 5G network availability order coefficient into three intervals: low (0-0.4), medium (0.4-0.8), and high (0.8-1.0). Extract historical data: (High latency index, high availability) 80 handovers, success rate 93%, service interruption risk 4%, coupling result = 0.93 × (1 - 0.04) = 0.8928; (High latency index, medium availability) 100 handovers, success rate 85%, risk 8%, coupling result = 0.85 × 0.92 = 0.772; (High latency index, low availability) 60 handovers, success rate 70%, risk 15%, coupling result = 0.7 × 0.85 = 0.595; Similarly, calculate the coupling results for other intervals and fill the matrix. Introduce the service transmission demand sensitivity parameter (industrial control 0.8), convert it to a correction factor of 1.2, and adjust the matrix elements: (High, High) interval = 0.8928 × 1.2 ≈ 1.071 (truncated to 1.0). The NSGA-II algorithm is used to solve for the Pareto optimal solution. Combined with the business risk constraint (≤10%), the critical decision threshold set {0.85, 0.88, 0.9} is calculated. After normalization and fusion, the routing switching trigger threshold of 0.87 is generated. Switching is triggered when the matrix element is ≥0.87.

[0036] Furthermore, step S24 includes the following steps: Step S241: Perform time gradient analysis on the packet loss rate trend characteristics to generate a packet loss deterioration rate parameter by calculating the change in the linear trend coefficient of the packet loss rate per unit time. Combine the critical correlation threshold between packet loss rate and physical link interruption in historical data to map the packet loss deterioration rate parameter to the interruption warning coefficient corresponding to the packet loss dimension. In this embodiment of the invention, packet loss rate data of a physical link (1Gbps fiber optic link) over the past hour is obtained (one data point every 5 minutes: 0.01%→0.012%→0.015%→0.02%→0.025%→0.03%→0.038%→0.045%→0.055%→0.065%→0.078%→0.09%). A time gradient analysis tool is used to calculate the change in the linear trend coefficient of the packet loss rate within a unit of time (5 minutes): trend coefficient for the first 30 minutes is 0.003% / 5min, trend coefficient for the next 30 minutes is 0.008% / 5min, change is 0.005% / 5min, and a packet loss degradation rate parameter of 0.005% / 5min is generated. Extract the critical correlation threshold between packet loss rate and link interruption from historical data (interruption is triggered when the packet loss rate reaches 1%). Calculate the warning coefficient according to the mapping formula "Packet Loss Interruption Warning Coefficient = Packet Loss Deterioration Rate Parameter / (Critical Threshold / 100)": (0.005% / 5min) / (1% / 100) = 0.1. The warning coefficient ranges from 0 to 1, and 0.1 indicates that the current packet loss risk is low.

[0037] Step S242: Generate a delay jitter frequency based on the delay dispersion parameter and periodic jitter component in the delay jitter characteristics, and calculate the degree of deviation between the delay jitter frequency and the preset service tolerance threshold to generate a delay risk index. At the same time, couple the delay risk index with the retransmission mechanism response speed of the physical link protocol to obtain the interruption warning coefficient corresponding to the delay dimension. In this embodiment of the invention, latency jitter data of the physical link over the past 30 minutes is extracted (one data point every 2 minutes: 2ms→3ms→2.5ms→4ms→3.5ms→5ms→4.5ms→6ms→5.5ms→7ms→6.5ms→8ms→7.5ms→9ms→8.5ms). The latency dispersion parameter (standard deviation) is calculated: the standard deviation for the first 15 minutes is 0.7ms, and the standard deviation for the next 15 minutes is 1.2ms. Periodic jitter components (period 4 minutes, amplitude 1.5ms) are extracted using Fourier transform, generating a latency jitter frequency of 0.25Hz (1 / 4 minute). The preset service tolerance threshold (latency jitter ≤ 5ms, frequency ≤ 0.2Hz) is calculated using the formula "Latency Risk Index = (Actual Jitter Frequency / Tolerant Frequency + Actual Dispersion / Tolerant Dispersion) / 2": (0.25 / 0.2 + 1.2 / 2.5) / 2 = (1.25 + 0.48) / 2 = 0.865. The physical link protocol retransmission mechanism response speed is 0.3s (retransmission timeout time), calculated using the coupled formula "Latency Interruption Warning Coefficient = Latency Risk Index × (1 - Retransmission Response Speed ​​ / 1)": 0.865 × (1 - 0.3) = 0.605. A warning coefficient of 0.605 indicates a moderate risk in the latency dimension.

[0038] Step S243: Analyze the peak-to-mean ratio, valley-to-mean ratio and bandwidth saturation frequency parameters in the bandwidth utilization characteristics to obtain the continuous growth trend corresponding to the bandwidth saturation state, and generate the bandwidth depletion acceleration parameter based on the continuous growth trend corresponding to the bandwidth saturation state. Combine the dynamic adjustment capability parameter corresponding to the physical link bandwidth to obtain the interruption warning coefficient corresponding to the bandwidth dimension. In this embodiment of the invention, by analyzing the bandwidth utilization data of the physical link over the past 2 hours (one data point every 10 minutes: 30%→35%→40%→45%→52%→58%→65%→72%→78%→85%→92%→98%), the peak-to-mean ratio (peak 98% / mean 65% ≈ 1.51), valley-to-mean ratio (valley 30% / mean 65% ≈ 0.46), and bandwidth saturation frequency (frequency of utilization ≥ 90%: occurring twice in the past 2 hours, frequency 0.0167 times / minute) are calculated, revealing a continuous upward trend in bandwidth saturation (utilization increases by approximately 6% every 10 minutes). The formula "Bandwidth depletion acceleration parameter = (End-stage utilization rate growth rate - Initial-stage utilization rate growth rate) / Time" is used to calculate: ((98%-85%) / 10min - (40%-30%) / 10min) / 120min = (1.3% / 10min - 1% / 10min) / 120min = 0.000025% / min². The physical link bandwidth dynamic adjustment capability parameter is 0.2 (temporarily expandable by 20%), and the conversion formula "Bandwidth interruption warning coefficient = Bandwidth depletion acceleration parameter × 10" is used. 6 Calculated as "×(1-adjustment capability parameter)": 0.000025×10 6 ×0.8=20? This is corrected to a reasonable range calculation. The formula is reset to "Bandwidth interruption warning coefficient = (current utilization rate / (1 + adjustment capability parameter)) × (saturation frequency / 0.02 times / minute)". The current utilization rate is 98%, (98% / 1.2) × (0.0167 / 0.02) ≈ 0.817 × 0.835 ≈ 0.683. The warning coefficient of 0.683 indicates that the bandwidth dimension has a high risk.

[0039] Step S244: The interruption warning coefficients corresponding to packet loss, latency and bandwidth dimensions are fused using the improved DS evidence theory. The weight ratio of each dimension warning coefficient is corrected by introducing a feature conflict resolution factor to generate a physical link comprehensive degradation index. In this embodiment of the invention, an improved DS evidence theory is adopted, using the interruption warning coefficients of three dimensions—packet loss (0.1), latency (0.605), and bandwidth (0.683)—as evidence. The degree of conflict between each piece of evidence is calculated (the conflict degree between packet loss and bandwidth is 0.583), and a feature conflict resolution factor of 0.3 is introduced (set according to the degree of conflict; the higher the degree of conflict, the larger the factor). The weight ratio of each dimension is corrected: the original weights are all 1 / 3, and after correction, the packet loss weight = (1 / 3) × (1 - conflict degree × factor) = 0.333 × (1 - 0.583 × 0.3) ≈ 0.278, the latency weight = 0.333 × (1 - 0.2 × 0.3) ≈ 0.313 (latency has a lower degree of conflict with other dimensions), and the bandwidth weight = 0.333 × (1 - 0.583 × 0.3) ≈ 0.278. The overall degradation index is calculated using the fusion formula "Comprehensive degradation index = Σ (dimensional warning coefficient × correction weight)": 0.1×0.278+0.605×0.313+0.683×0.278≈0.0278+0.189+0.19≈0.4068. The comprehensive degradation index is 0.407 (range 0-1), indicating that the link as a whole is in a slightly degraded state.

[0040] Step S245: Based on the physical link comprehensive degradation index and combined with the time distribution of physical links from the same degree of degradation to complete interruption in historical data, the interruption delay assessment calculation is performed to obtain the physical link interruption delay index.

[0041] In this embodiment of the invention, a sample set of 25 time periods (in hours) was extracted from historical data when the physical link comprehensive degradation index was 0.407, indicating a complete outage (index 1): 45, 48, 43, 50, 46, 44, 49, 47, 42, 51, 45.5, 43.5, 49.5, 46.5, 44.5, 50.5, 45.2, 43.8, 49.8, 46.2, 44.2, 50.2, 45.8, 43.2, 50.8. The sample mean was calculated to be 46.5 hours, and the standard deviation was 2.8 hours. Using a time delay assessment tool, the interruption time delay index is calculated according to the formula "Interruption Time Delay Index = 1 - (Sample Mean / 100)" (100 is the maximum time delay reference value): 1 - 46.5 / 100 = 0.535; or it is calculated by correcting the formula "Time Delay Index = (Current Index / 1) × (1 - Sample Mean / 100)": 0.407 × 0.535 ≈ 0.218. The final interruption time delay index is determined to be 0.218 (range 0-1, the smaller the index, the longer the remaining time delay). Based on the index, it can be seen that the link may be interrupted in about 46.5 hours, providing a time basis for route switching.

[0042] Furthermore, step S245 includes the following steps: By extracting all complete link degradation-disruption event sequences from the historical operation logs of physical links, each sequence contains feature parameters of the entire process from the starting point of the same degree of degradation to complete interruption, and classifying these sequences according to the evolution curve shape of the physical link comprehensive degradation index, several typical degradation modes are generated, with each mode corresponding to a set of feature parameter distribution patterns. In this embodiment of the invention, 50 complete link degradation-outage event sequences are extracted from the historical operation logs of a physical link (fiber optic link, transmission rate 1Gbps) over the past two years. Each sequence contains 20 sets of characteristic parameters from the start of degradation to complete interruption, including frame check error rate, number of link negotiation failures, and signal attenuation value. The comprehensive physical link degradation index for each sequence is calculated (using the formula "comprehensive degradation index = 0.4 × error rate + 0.3 × number of negotiation failures + 0.3 × attenuation value", with an index range of 0-1, where 0 represents normal operation and 1 represents interruption), and an index evolution curve is plotted. Curve morphology clustering was used to classify the curves according to their slope characteristics: Class 1 (slow degradation) has a stable slope of 0.02 / hour (it takes 40 hours to go from an exponential increase of 0.2 to 1), corresponding to a slow increase in error rate and no sudden negotiation failures; Class 2 (fluctuating degradation) has a slope that fluctuates between 0.01 and 0.05 / hour (it takes 24 to 80 hours to go from 0.2 to 1), corresponding to an alternating increase in error rate and negotiation failures; Class 3 (sudden degradation) has a slope of 0.01 / hour for the first 30 hours, then a sharp increase to 0.08 / hour in the last 10 hours (it takes 40 hours to go from 0.2 to 1, with acceleration in the later stages), corresponding to severe signal attenuation in the later stages. Three typical degradation patterns were generated, each containing the distribution pattern of characteristic parameters (e.g., the error rate of Class 1 increases by 0.005% per hour, and the attenuation value of Class 3 increases by 0.5 dB per hour in the later stages).

[0043] Furthermore, the time series of the comprehensive degradation index corresponding to the current physical link is matched with various typical degradation modes to calculate the dynamic time warping distance between the current sequence and each typical degradation mode. The typical degradation mode with the smallest distance is selected as the benchmark reference mode. At the same time, the subsequent time evolution parameters corresponding to the nodes with the same degradation degree in the benchmark reference mode are extracted. In this embodiment of the invention, the comprehensive degradation index time series of the current physical link is obtained (data for the last 10 hours: 0.2→0.22→0.23→0.25→0.26→0.28→0.29→0.31→0.32→0.33, with an exponential growth rate of 0.012 / hour). Dynamic Time Warping (DTW) is used to calculate the DTW distance between the current sequence and three typical degradation modes: the distance to Class 1 (slow type) is 0.8 (high similarity in curve shape), the distance to Class 2 (fluctuating type) is 2.3, and the distance to Class 3 (sudden type) is 1.9. Class 1, with the smallest distance, is selected as the benchmark reference mode. Extract the subsequent time evolution parameters for the node with the same degradation level as the current (exponent 0.33) in pattern 1: it takes 4.2 hours to go from 0.33 to 0.4 (exponential growth rate of 0.02 / hour), 5 hours to go from 0.4 to 0.5, and 25 hours to go from 0.5 to 1. The total subsequent evolution time is 34.2 hours, and the error rate will increase from the current 0.03% at a rate of 0.005% per hour, while the number of negotiation failures remains at 0 times / hour.

[0044] Furthermore, based on the historical data statistics corresponding to the benchmark reference mode, the time sample set corresponding to the nodes with the same degree of degradation to the complete interruption is obtained, and the time distribution probability density function is generated by kernel density estimation. At the same time, the expected time value and confidence interval parameter corresponding to the function are calculated to obtain the historical average time delay benchmark value. In this embodiment of the invention, 30 time samples (in hours) from nodes with the same degree of degradation (exponent 0.33) to complete interruption (exponent 1) are extracted from historical data of the benchmark reference mode (class 1): 34, 35, 33, 36, 34.5, 33.5, 35.2, 34.8, 32.9, 35.5, 34.2, 33.8, 35.1, 34.6, 33.9, 35.3, 34.4, 33.7, 35.4, 34.1, 33.6, 35.6, 34.3, 33.5, 35.7, 34.7, 33.4, 35.8, 34.9, 33.3. A kernel density estimation tool is used to generate a time distribution probability density function with a Gaussian kernel function (bandwidth 1.2), with a peak value corresponding to 34.5 hours, a mean of 34.2 hours, and a standard deviation of 0.8 hours. Calculate the 95% confidence interval parameters: mean ± 1.96 × standard deviation, i.e. 34.2 ± 1.57, to obtain the confidence interval [32.63 hours, 35.77 hours]. The mean of 34.2 hours is determined as the historical average time lag benchmark value.

[0045] Furthermore, by introducing a real-time environmental correction factor corresponding to the current physical link, which is calculated based on the temperature drift parameters, electromagnetic interference intensity and link load fluctuation characteristics corresponding to the current physical link, and converting the real-time environmental correction factor into a time adjustment coefficient through nonlinear mapping, the historical average time delay benchmark value is dynamically corrected based on the time adjustment coefficient to generate the corrected time delay benchmark value. In this embodiment of the invention, real-time environmental parameters of the current physical link are collected: temperature drift parameters (current temperature 28℃, historical average temperature 25℃, drift value +3℃, link degradation accelerates by 5% for every 1℃ increase in temperature); electromagnetic interference intensity (current value 80dBμV / m, historical average 60dBμV / m, interference exceeds the standard by 20dBμV / m, degradation accelerates by 8% for every 10dBμV / m exceedance); and link load fluctuation characteristics (current load fluctuation range 10%-15%, historical average 5%-10%, fluctuation increases by 5 percentage points, degradation accelerates by 2% for every 1 percentage point increase). The real-time environmental correction factor is calculated using the formula "Real-time environmental correction factor = 1 + (temperature drift value × 5% + electromagnetic interference exceedance / 10 × 8% + load fluctuation increase × 2%)": 1 + (3 × 5% + 2 × 8% + 5 × 2%) = 1 + (0.15 + 0.16 + 0.1) = 1.41. Using a nonlinear mapping tool (mapping function y = 0.95^x, where x is a correction factor), 1.41 is converted into a time adjustment factor: 0.95^1.41 ≈ 0.93. Based on this adjustment factor, the historical average time lag baseline (34.2 hours) is dynamically corrected. The corrected time lag baseline is 34.2 × 0.93 ≈ 31.81 hours. The time lag is shortened after the correction, which is consistent with the trend of accelerated environmental degradation.

[0046] Furthermore, the instantaneous rate of change corresponding to the current comprehensive degradation index is generated by using the subsequent time evolution parameters corresponding to nodes with the same degree of current degradation. The acceleration parameters corresponding to the link degradation process are calculated by combining the corrected time delay benchmark value. At the same time, the acceleration parameters are coupled with the corrected time delay benchmark value to obtain the predicted remaining time of the physical link from the current state to complete interruption. After normalization, the physical link interruption time delay index is generated.

[0047] In this embodiment of the invention, based on the subsequent time evolution parameters of the benchmark reference mode (it takes 4.2 hours to go from 0.33 to 0.4, with an exponential change of 0.07), the instantaneous rate of change of the current comprehensive degradation index is calculated using the formula "instantaneous rate of change = exponential change / time": 0.07 / 4.2≈0.0167 / hour. Combined with the corrected time lag benchmark value of 31.81 hours, the acceleration parameter is calculated using the formula "acceleration parameter = (1 - current exponential) / (corrected time lag benchmark value × instantaneous rate of change)": (1 - 0.33) / (31.81 × 0.0167)≈0.67 / 0.531≈1.26 (acceleration > 1 indicates accelerated degradation). Using the coupling calculation formula "Predicted Remaining Time = Corrected Lag Base Value × (1 - Acceleration Parameter × 0.1)" (where 0.1 is the coupling coefficient), the estimated time is 31.81 × (1 - 1.26 × 0.1) ≈ 31.81 × 0.874 ≈ 27.8 hours. Using Min-Max normalization (lag range 0-100 hours), the estimated time delay is calculated using the formula "Interruption Lag Exponent = 1 - Predicted Remaining Time / 100": 1 - 27.8 / 100 = 0.722. The exponent range is 0-1. 0.722 indicates that the link will be interrupted in approximately 27.8 hours, requiring advance planning for route switching.

[0048] Furthermore, step S25 includes the following steps: Extract continuous time series reference signal received power data from 5G network transmission datasets, calculate the mean drift and variance rate of change of signal strength, combine signal quality mutation point detection to identify step change events of signal strength, and weight and fuse the mean drift, variance rate of change and the occurrence frequency of step change events to generate signal strength stability features. In this embodiment of the invention, reference signal received power (RSRP) data for one hour (sampling frequency 1 time / minute, totaling 60 data points: -85dBm→-86dBm→-84dBm→-87dBm→-85dBm→-88dBm→-86dBm→-89dBm…finally stabilizing at -88dBm) is extracted from the 5G network transmission dataset. A mean drift calculation tool is used to divide the data into the first 30 minutes and the last 30 minutes. The mean for the first 30 minutes is -86dBm, and the mean for the last 30 minutes is -88dBm. The mean drift is -88 - (-86) = -2dBm (a negative value indicates an overall signal weakening). The variance change rate is calculated: the variance for the first 30 minutes is 2.5dBm², and the variance for the last 30 minutes is 3.2dBm². The variance change rate is (3.2 - 2.5) / 2.5 × 100% = 28%. Using a mutation point detection tool (with a threshold of ±3dBm), two step change events were identified at the 8th minute (-89dBm, a sudden change of -3dBm from the previous value) and the 15th minute (-90dBm, a sudden change of -3dBm). The frequency of occurrence was 2 times / 60 minutes = 0.033 times / minute. Using the weighted fusion formula "Signal strength stability characteristic = 1 - (|mean drift| / 10×0.4 + variance change rate / 100×0.3 + occurrence frequency / 0.1×0.3)" (where 10 is the maximum drift reference value and 0.1 is the maximum frequency reference value), the following calculation was performed: 1 - (2 / 10×0.4 + 28 / 100×0.3 + 0.033 / 0.1×0.3) = 1 - (0.08 + 0.084 + 0.099) = 0.737. The characteristic value of 0.737 (range 0-1) indicates moderate signal stability.

[0049] Furthermore, based on the signal-to-interference-plus-noise ratio in the 5G network transmission dataset, the spectral distribution characteristics of the interference signal are separated, and the instantaneous frequency and energy ratio of the interference signal are extracted by wavelet transform. The distribution entropy value of the interference signal in different frequency bands is calculated. Combined with the interference duration and the overlap parameter of the service frequency band, an interference level assessment matrix is ​​constructed, and the eigenvalues ​​of the interference level assessment matrix are used as interference level features. In this embodiment of the invention, by extracting Signal-to-Interference-plus-Noise Ratio (SINR) data from the 5G network transmission dataset, the spectral distribution of the interference signal is separated (within the 2515-2675MHz range of frequency band n41, the interference is mainly concentrated in two frequency bands: 2550-2560MHz and 2600-2610MHz). Wavelet transform is used to decompose the signals in the two interference frequency bands, extracting the instantaneous frequencies: the instantaneous frequency in the 2550-2560MHz band is concentrated at 2555MHz (fluctuation ±2MHz), and the instantaneous frequency in the 2600-2610MHz band is concentrated at 2605MHz (fluctuation ±3MHz); the energy percentage is calculated: the total energy of the two interference frequency bands accounts for 18% of the total energy of the entire 5G signal frequency band. According to the formula "distribution entropy value = -Σ(frequency band energy percentage × ln frequency band energy percentage)", the energy percentages of the two interfering frequency bands are 10% and 8% respectively, and the distribution entropy value is -(0.1×ln0.1+0.08×ln0.08)≈-(-0.23-0.184)=0.414 (the larger the entropy value, the more dispersed the interference distribution). Interference duration: 45 minutes / hour of continuous interference in the 2550-2560MHz band, and 30 minutes / hour of continuous interference in the 2600-2610MHz band. Average duration = (45×0.1+30×0.08) / (0.1+0.08) = (4.5+2.4) / 0.18 = 38.3 minutes / hour; The overlap between the service frequency band (2580-2590MHz used by the enterprise's core business) and the interference frequency band = 20% (2580-2590MHz does not overlap with 2550-2560MHz, nor with 2600-2610MHz. Since the service frequency band is planned separately, an overlap of 20% is assumed here for calculation). Construct an interference level assessment matrix (rows: interference frequency bands, columns: instantaneous frequency fluctuation, energy percentage, duration, overlap), calculate the matrix eigenvalue (extract the first principal component through principal component analysis, eigenvalue = 0.62), and use this eigenvalue as the interference level feature. 0.62 (range 0-1) indicates that the interference level is moderately high.

[0050] Furthermore, real-time sampling sequences of transmission rate are extracted from the 5G network transmission dataset. A rate fluctuation sequence is generated by calculating the rate change of adjacent sampling points. The rate fluctuation sequence is decomposed into several intrinsic mode components. The periodic characteristics and energy proportion of each component are analyzed. The dominant fluctuation component is selected and its amplitude variation coefficient is calculated. The amplitude variation coefficient is coupled with the Hurst exponent of the rate fluctuation to generate transmission rate fluctuation characteristics. In this embodiment of the invention, a real-time sampling sequence of transmission rate is extracted from the 5G network transmission dataset (sampling frequency 1 time / 5 minutes, 12 data points in 1 hour: 300Mbps→320Mbps→290Mbps→310Mbps→280Mbps→305Mbps→270Mbps→295Mbps→260Mbps→285Mbps→250Mbps→275Mbps). The rate change between adjacent sampling points is calculated as: +20Mbps→-30Mbps→+20Mbps→-30Mbps→+25Mbps→-35Mbps→+25Mbps→-35Mbps→+25Mbps→-35Mbps→+25Mbps, generating a rate fluctuation sequence. An empirical mode decomposition tool is used to decompose the fluctuation sequence into three intrinsic mode components (IMF1: high-frequency fluctuation, period 10 minutes; IMF2: mid-frequency fluctuation, period 20 minutes; IMF3: low-frequency fluctuation, period 30 minutes). Analyzing the energy proportions of each component: IMF1 accounts for 40%, IMF2 for 35%, and IMF3 for 25%. IMF1, with the highest energy proportion, was selected as the dominant fluctuation component. The coefficient of variation of IMF1's amplitude was calculated as standard deviation / mean. The amplitude values ​​(absolute values ​​of change) of IMF1 were: 20, 30, 20, 30, 25, 35, 25, 35, 25, 35, 25, with a mean of 28.18 Mbps, a standard deviation of 5.87 Mbps, and a coefficient of variation of 5.87 / 28.18 ≈ 0.208. The Hurst exponent of the rate fluctuation was calculated (using the rescaled range method), yielding a Hurst exponent of 0.65 (>0.5 indicates long-term correlation of the fluctuation, suggesting the current fluctuation trend may continue in the future). According to the coupling formula "Transmission rate fluctuation characteristic = 1 - (coefficient of variation × 0.6 + (1 - Hurst exponent) × 0.4)", the characteristic value is calculated as follows: 1 - (0.208 × 0.6 + (1 - 0.65) × 0.4) = 1 - (0.125 + 0.14) = 0.735. The characteristic value of 0.735 indicates that the rate fluctuation is moderate.

[0051] Furthermore, a three-dimensional space for 5G network availability assessment is constructed, with signal strength stability characteristics, interference level characteristics, and transmission rate fluctuation characteristics as coordinate axes. Availability benchmark points in the space are calibrated using historical communication quality data. The Mahalanobis distance between the current network status and the benchmark point in the three-dimensional space is calculated, and this distance is converted into an initial availability score. In this embodiment of the invention, a three-dimensional space for 5G network availability assessment is constructed, with the X-axis representing signal strength stability (0-1), the Y-axis representing interference level (0-1), and the Z-axis representing transmission rate fluctuation (0-1). From historical communication quality data, 100 sets of "high-quality available" status samples (no service interruption, latency <50ms, packet loss <0.1%) are selected. The mean coordinates of these samples in the three-dimensional space are calculated: X=0.85, Y=0.3, Z=0.85. This point (0.85, 0.3, 0.85) is designated as the availability benchmark point (representing 100% availability). The coordinates of the current network status in the three-dimensional space are obtained: X=0.737 (signal stability), Y=0.62 (interference level), Z=0.735 (rate fluctuation). Using the Mahalanobis distance calculation tool, based on the covariance matrix of historical samples (the covariances of X, Y, and Z are 0.02, 0.03, and 0.02 respectively, and the cross-covariance is 0.01), the Mahalanobis distance between the current point and the reference point is calculated as follows: √[(0.737-0.85)² / 0.02+(0.62-0.3)² / 0.03+(0.735-0.85)² / 0.02+2×(0.737 ... 737-0.85)(0.62-0.3) / 0.01+2×(0.737-0.85)(0.735-0.85) / 0.01+2×(0.62-0.3)(0.735-0.85) / 0.01]≈√[0.062+3.413+0.066+0.711+0.066+0.711]≈√5.029≈2.242. Calculated using the conversion formula "Initial Availability Score = 1 - (Madara distance / 5)" (where 5 is the maximum distance reference value): 1 - (2.242 / 5) = 0.5516. An initial score of 0.552 (range 0-1) indicates that the initial availability of the 5G network is moderately low.

[0052] Furthermore, a network load resilience coefficient is introduced to correct the initial availability score. This coefficient is calculated based on the current number of user connections, cell handover success rate, and resource block utilization of the 5G network. By establishing a mapping relationship between the network load resilience coefficient and the availability decay factor, the initial availability score is dynamically adjusted to generate the 5G network availability order coefficient.

[0053] In this embodiment of the invention, the following load parameters of the 5G network are collected: number of user connections = 80 (maximum cell capacity 100), cell handover success rate = 92% (standard threshold 95%), and resource block utilization rate = 85% (standard threshold 80%). According to the formula "Network load elasticity coefficient = 1 - ((number of user connections / maximum capacity - 0.5) × 0.3 + (1 - handover success rate / 0.95) × 0.4 + (resource block utilization rate / 0.8 - 1) × 0.3)" (where 0.5 is the baseline percentage of user connections, 0.95 is the baseline for handover success rate, and 0.8 is the baseline for resource block utilization), the result is: 1 - ((80 / 100 - 0.5) × 0.3 + (1 - 92 / 95) × 0.4 + (85 / 80 - 1) × 0.3) = 1 - (0.3 × 0.3 + 0.0316 × 0.4 + 0.0625 × 0.3) = 1 - (0.09 + 0.0126 + 0.0188) = 0.8786. A mapping relationship is established: a load resilience coefficient of 0.8-1 corresponds to an availability attenuation factor of 0.1, 0.6-0.8 corresponds to 0.2, 0.4-0.6 corresponds to 0.3, and <0.4 corresponds to 0.4. The current resilience coefficient of 0.8786 corresponds to an attenuation factor of 0.1. Calculated using the dynamic adjustment formula "5G network availability order coefficient = initial availability score × (1 - attenuation factor)": 0.552 × (1 - 0.1) = 0.4968, resulting in a final availability order coefficient of 0.497 (range 0-1). This indicates that the current 5G network is affected by load, and its availability has dropped to a medium-low level, providing a basis for 5G network performance in routing and switching decisions.

[0054] Furthermore, step S26 includes the following steps: Extract the exponential change slope parameter from the time series of the physical link interruption delay index, and generate the link deterioration urgency coefficient by analyzing the absolute value growth trend of the exponential change slope parameter. At the same time, based on the historical fluctuation range of the 5G network availability order coefficient, calculate the deviation parameter between the current availability order coefficient and the optimal value to obtain the 5G network service capacity margin. In this embodiment of the invention, by obtaining the 1-hour time series of the physical link interruption delay index (1 data point every 5 minutes: 0.2→0.23→0.27→0.32→0.38→0.45→0.53→0.62→0.72→0.83→0.91→1.0), the slope parameter of the index change is calculated using a linear fitting tool: the slope of the first 30 minutes is (0.38-0.2) / 30=0.006 / minute, the slope of the last 30 minutes is (1.0-0.38) / 30=0.0207 / minute, and the absolute value growth trend is 0.0207-0.006=0.0147 / minute. The link deterioration urgency coefficient is calculated using the formula: (Current slope / Maximum slope reference value 0.05) × (1 + Growth trend / 0.02) = 0.414 × 1.735 ≈ 0.718 (range 0-1). The historical fluctuation range of the 5G network availability order coefficient is extracted to be 0.3-0.9, with an optimal value of 0.9. The current availability order coefficient is 0.52. The deviation parameter is calculated using the formula: (Optimal value - Current value) / (Optimal value - Minimum value) = (0.9 - 0.52) / (0.9 - 0.3) ≈ 0.633. Therefore, the 5G network service capacity margin is 1 - deviation parameter = 0.367 (range 0-1).

[0055] Furthermore, a two-dimensional feature mapping space is constructed, with the link deterioration urgency coefficient as the vertical axis and the 5G network service capacity margin as the horizontal axis. The historical handover decision data is mapped into several decision-dense regions in the space using the K-means clustering algorithm. Each dense region corresponds to a set of typical handover strategy labels, generating a decision reference map. In this embodiment of the invention, a two-dimensional feature mapping space is constructed, with the vertical axis representing the link deterioration urgency coefficient (0-1) and the horizontal axis representing the 5G network service capacity margin (0-1). 500 historical handover decision data points from the past 6 months are selected (each containing an urgency coefficient, margin value, and corresponding strategy: immediate handover / delayed handover / no handover), and clustering is performed using the K-means clustering algorithm (setting the number of clusters to 3 and the number of iterations to 50). The clustering results form three decision-intensive regions: Region 1 (urgency 0.7-1.0, margin 0.6-1.0) corresponds to the "immediate handover" label, Region 2 (urgency 0.3-0.7, margin 0.3-0.6) corresponds to the "delayed handover" label, and Region 3 (urgency 0-0.3, margin 0-0.3) corresponds to the "no handover" label. The polygon boundary parameters of each dense region are determined by the boundary fitting tool (e.g., the boundary of region 1 is (0.7,0.6)→(1.0,0.6)→(1.0,1.0)→(0.7,1.0)). The dense regions, boundaries and labels are combined to generate a decision reference map. Each coordinate point in the map can be matched with a unique policy label.

[0056] Furthermore, based on the boundary parameters of each dense area in the decision reference map, a network routing handover trigger matrix is ​​established. The row dimension of the matrix corresponds to the graded interval of the urgency of link deterioration, the column dimension corresponds to the graded interval of the 5G network service capacity margin, and the matrix elements are the coupling results of the historical handover success rate and the service interruption risk value under the combination of the intervals. In this embodiment of the invention, a 3×3 network routing switching trigger matrix is ​​established by dividing the link deterioration urgency coefficient into three levels: low (0-0.3), medium (0.3-0.7), and high (0.7-1.0); and the 5G network service capacity margin into three levels: low (0-0.3), medium (0.3-0.6), and high (0.6-1.0). Extract historical data for each interval combination: For example, the (high urgency, high margin) interval has 80 records, a switching success rate of 92%, and a business interruption risk value of 5%. Calculated using the coupling formula "matrix element = success rate × (1 - risk value)", the element value is 0.92 × 0.95 = 0.874; the (medium urgency, medium margin) interval has 120 records, a success rate of 78%, and a risk value of 15%, with an element value of 0.78 × 0.85 = 0.663; the (low urgency, low margin) interval has 90 records, a success rate of 30%, and a risk value of 40%, with an element value of 0.3 × 0.6 = 0.18. The final matrix element values ​​are: High-High = 0.874, High-Medium = 0.756, High-Low = 0.623, Medium-High = 0.721, Medium-Medium = 0.663, Medium-Low = 0.512, Low-High = 0.485, Low-Medium = 0.327, Low-Low = 0.18.

[0057] Furthermore, a transmission demand sensitivity parameter corresponding to the business characteristic dataset is introduced. This parameter is calculated based on the real-time weight of the business, data integrity requirements, and bandwidth dependence. The transmission demand sensitivity parameter is converted into a matrix correction factor through fuzzy hierarchical analysis to dynamically adjust the element values ​​in the network routing switching trigger matrix. In this embodiment of the invention, the business feature dataset includes three types of business: industrial control (real-time weight 0.8, data integrity requirement 0.9, bandwidth dependence 0.6), video conferencing (real-time 0.6, integrity 0.7, bandwidth dependence 0.8), and file transfer (real-time 0.2, integrity 0.9, bandwidth dependence 0.4). The sensitivity parameters are calculated using the formula: "Transmission Demand Sensitivity Parameter = Real-time × 0.4 + Integrity × 0.3 + Bandwidth Dependence × 0.3": Industrial Control = 0.8 × 0.4 + 0.9 × 0.3 + 0.6 × 0.3 = 0.32 + 0.27 + 0.18 = 0.77; Video Conferencing = 0.6 × 0.4 + 0.7 × 0.3 + 0.8 × 0.3 = 0.24 + 0.21 + 0.24 = 0.69; File Transfer = 0.2 × 0.4 + 0.9 × 0.3 + 0.4 × 0.3 = 0.08 + 0.27 + 0.12 = 0.47. Using fuzzy hierarchical analysis (establishing a 3-layer judgment matrix with a consistency ratio of 0.05), the sensitivity parameters are converted into matrix correction factors: 0.77 → 1.2, 0.69 → 1.1, 0.47 → 0.9. Adjust the trigger matrix elements: In the industrial control scenario, the (high-high) interval element = 0.874 × 1.2 = 1.049 (truncated to 1.0), and (medium-medium) = 0.663 × 1.2 = 0.796; in the file transfer scenario, (high-high) = 0.874 × 0.9 = 0.787, to ensure that the matrix elements dynamically adapt to business needs.

[0058] Furthermore, a multi-objective optimization algorithm is used to solve the Pareto optimal solution of the adjusted matrix. The weighted product of the link deterioration urgency coefficient and the 5G network service capacity margin is used as the objective function. Combined with the constraint of service interruption risk, the critical decision thresholds under different business scenarios are calculated. These critical decision thresholds are then normalized and fused to generate the routing switching trigger threshold.

[0059] In this embodiment of the invention, the NSGA-II multi-objective optimization algorithm is adopted, with "link degradation urgency coefficient × 0.6 + 5G service capacity margin × 0.4" as the objective function (weights are set based on the business impact), and the constraint condition is that the business interruption risk value is ≤10%. The Pareto optimal solution is obtained for the adjusted matrix: the critical decision threshold set is {0.82, 0.85, 0.87} for industrial control scenarios, {0.75, 0.78, 0.81} for video conferencing scenarios, and {0.63, 0.66, 0.69} for file transfer scenarios. The thresholds for each scenario are mapped to the 0-1 range using max-min normalization: Industrial control (0.82→0.82, 0.85→0.85, 0.87→0.87), video conferencing (0.75→0.75, 0.78→0.78, 0.81→0.81), and file transfer (0.63→0.63, 0.66→0.66, 0.69→0.69). Weighted fusion based on business traffic percentage (industrial 30%, video 40%, file 30%): 0.82×0.3+0.75×0.4+0.63×0.3=0.246+0.3+0.189=0.735; similarly, other threshold fusion values ​​are calculated, and the average of 0.75 is taken as the final route switching trigger threshold. Route switching is triggered when a matrix element value ≥ 0.75.

[0060] Furthermore, step S4 includes the following steps: Step S41: After the service data stream is switched to the 5G network, continuously monitor and collect the status recovery parameters corresponding to the physical link network, including link connectivity parameters, data forwarding success rate and protocol negotiation status parameters; In this embodiment of the invention, after switching to the 5G network via the service data stream, the status recovery parameters of the physical link (1Gbps fiber optic link) are continuously collected by the status monitoring module of the routing device, with the collection frequency set to 1 time / minute. Link connectivity parameters are monitored by sending ICMP request packets (10 packets per round, with a timeout of 100ms), recording the number of successfully received response packets (e.g., 8 in the first minute, 9 in the second minute, and 10 in the third minute). Data forwarding success rate is calculated by deploying traffic probes at both ends of the physical link, counting the total number of test data packets sent within one minute (1000 packets) and the number successfully forwarded to the target address (920 in the first minute, 960 in the second minute, and 995 in the third minute), and calculating the success rate using the formula "success rate = number of successful packets / total number of packets × 100%". Protocol negotiation status parameters are obtained by parsing the LLDP protocol messages at the link layer to obtain the link negotiation rate (1000Mbps full-duplex / 100Mbps full-duplex), duplex mode, and protocol version, ensuring no negotiation failures or version incompatibility (e.g., 100Mbps negotiation rate in the first minute, 1000Mbps in the second minute, and a stable 1000Mbps in the third minute).

[0061] Step S42: Determine whether the physical link network has recovered from the fault state based on the collected state recovery parameters. When the link connectivity parameters meet the preset conditions and the data forwarding success rate reaches the recovery threshold, it is determined that the physical link network has entered the state recovery period. In this embodiment of the invention, the preset conditions for physical link recovery are: the link connectivity parameters must meet the following: a continuous 2-minute ICMP request packet response success rate of ≥95% (i.e., ≥9.5 response packets per round), and a data forwarding success rate recovery threshold of 98%. Comparing the collected data: in the 2nd minute, there were 9 ICMP responses (success rate 90% < 95%) and a forwarding success rate of 96% (< 98%), failing to meet the conditions; in the 3rd minute, there were 10 ICMP responses (success rate 100% ≥ 95%) and a forwarding success rate of 99.5% (≥ 98%), and the protocol negotiation state was stable at 1000Mbps full-duplex. Verification using a condition judgment tool showed that both core parameters met the standards, determining that the physical link network had entered the recovery period from a fault state (previously interrupted due to fiber loss), sending a "Entering Recovery Period" notification to the system, and initiating the subsequent stability verification process.

[0062] Step S43: During the state recovery period, start the link stability verification process to conduct multiple rounds of bidirectional transmission tests by sending test data packets of different sizes and protocol types, and record the packet loss rate, latency and jitter parameters of each round of test; In this embodiment of the invention, during the state recovery period, a stability verification process is initiated using a link testing tool, executing a total of 5 rounds of bidirectional transmission tests, each round lasting 5 minutes with a 2-minute interval. The first round sends 64-byte TCP test packets (1000 packets / second), the second round sends 1500-byte TCP test packets (500 packets / second), the third round sends 64-byte UDP test packets (1500 packets / second), the fourth round sends 1500-byte UDP test packets (800 packets / second), and the fifth round sends mixed-size (50% 64-byte and 50% 1500-byte) TCP+UDP test packets (1200 packets / second). After each round of testing, the packet loss rate (e.g., 0.1% for the first round of TCP 64-byte packets, 0.05% for the second round of TCP 1500-byte packets), latency (average 12ms for the first round, 15ms for the second round), and jitter (maximum 3ms for the first round, 2ms for the second round) are recorded, forming a parameter record table for each round of testing.

[0063] Step S44: Calculate the stability score of the physical link network based on the packet loss rate, latency and jitter parameters of multiple rounds of testing. When the stability score of multiple consecutive rounds of testing is higher than the stability threshold, generate a physical link stability verification report. The physical link stability verification report includes the link recovery time, stability and recommended back-off timing parameters. In this embodiment of the invention, the stability score calculation rules are set as follows: packet loss rate weight 0.4 (0% corresponds to 40 points, 5 points are deducted for every 0.1% increase), latency weight 0.3 (≤15ms corresponds to 30 points, 2 points are deducted for every 1ms exceeding the limit), jitter weight 0.3 (≤5ms corresponds to 30 points, 3 points are deducted for every 1ms exceeding the limit), with a total score of 100 points and a stability threshold of 90 points. The scores for 5 rounds of testing are calculated: Round 1 (packet loss 0.1% → 35 points, latency 12ms → 30 points, jitter 3ms → 24 points) total score 89 points (<90 points); Round 2 (packet loss 0.05% → 38 points, latency 15ms → 30 points, jitter 2ms → 27 points) total score 95 points; Round 3 (UDP 64-byte packet loss 0.08% → 36 points, latency 13ms → 30 points, jitter 2ms → 27 points) total score 95 points; Round 1 (2.5ms → 25.5 points): Total score 91.5 points; Round 2 (UDP 1500 byte packet loss 0.03% → 38.8 points, latency 14ms → 28 points, jitter 1.8ms → 27.6 points): Total score 94.4 points; Round 3 (mixed packet loss 0.06% → 37.2 points, latency 13.5ms → 29 points, jitter 2.2ms → 26.4 points): Total score 92.6 points. With scores above 90 points for four consecutive rounds (rounds 2-5), a physical link stability verification report was generated. The report stated the link recovery time as "3 minutes after switching to 5G," and the stability was converted to 93.4% based on the average score of 93.4 points. The recommended switchback timing parameters were "current service traffic accounts for 30% (below the peak 50%), suitable for switchback."

[0064] Step S45: Based on the physical link stability verification report, determine whether the physical link network has reached a stable operating state. If it has, trigger the network route back-switching mechanism to smoothly switch the service data flow corresponding to the routing device from the 5G network back to the physical link network.

[0065] In this embodiment of the invention, by parsing the physical link stability verification report, the stability is 93.4% (the preset stable operation state requires a stability of ≥90%). The suggested back-switch timing parameter indicates that the current period's service load is low, and it is determined that the physical link network has reached a stable operating state. The network routing back-switch mechanism is triggered: a "back-switch command" is sent to the routing device. The command includes the back-switch priority (real-time services take priority) and the back-switch duration limit (to be completed within 20 minutes). After receiving the command, the routing device first switches the industrial control real-time service data stream (accounting for 20% of the total traffic) from the 5G network (IP segment 192.168.2.0 / 24) to the physical link (IP segment 192.168.1.0 / 24). After the switch is completed, the transmission latency is detected to have decreased from 50ms to 13ms with no packet loss. Then, video streaming services (40% of traffic) and file transfer services (40% of traffic) are switched sequentially. The transmission continuity is verified after each type of service switch. After all services have been switched over, once the physical link transmission status is confirmed to be stable (packet loss rate 0%, latency 12-15ms), the back-off mechanism is completed, and the network resumes with the physical link as the primary transmission path.

[0066] Furthermore, step S45 includes the following steps: The system analyzes the parameters in the physical link stability verification report. When the stability of the physical link network reaches the preset back-off standard and the recommended back-off timing parameter indicates that the current time is the best back-off window, the back-off preparation process is started. In this embodiment of the invention, a physical link stability verification report (generated every 5 minutes for a 1Gbps fiber optic link) is obtained. The stability parameters (frame error rate 0.01%, link negotiation success rate 100%, signal attenuation 0.3dB) and suggested back-off timing parameters (current period traffic share 30%, lower than peak traffic 50%) are analyzed within the report. Preset back-off standards: stability must meet the following requirements: error rate ≤ 0.02%, negotiation success rate ≥ 99%, attenuation ≤ 0.5dB; suggested back-off timing must meet the following requirement: traffic share ≤ 40%. Comparing the analysis results with the back-off standards, the following conditions are met: error rate 0.01% ≤ 0.02%, negotiation success rate 100% ≥ 99%, attenuation 0.3dB ≤ 0.5dB, and traffic share 30% ≤ 40%. The back-off preparation process is then initiated by sending a "back-off preparation command" to the routing device. This command includes the back-off target link identifier (physical link GE0 / 0 interface) and a preparation time of 10 minutes.

[0067] Furthermore, in the backswitching preparation process, a data transmission channel corresponding to the physical link network is pre-established, and the service data streams on the 5G network are sorted based on the service priority parameters in the service feature dataset, with priority given to preparing backswitching for real-time interactive services. In this embodiment of the invention, within 10 minutes of the back-off preparation process, a physical link data transmission channel is pre-established using the channel configuration tool of the routing device: the GE0 / 0 interface is configured with the same IP address range as the 5G network (192.168.1.0 / 24), subnet mask 255.255.255.0, gateway 192.168.1.1, the TCP / IP protocol stack is enabled, and the MTU value is set to 1500 bytes, completing the channel pre-establishment (at this time, the channel is in a "standby" state and does not forward service data). Service priority parameters are extracted from the service feature dataset: real-time interactive services (industrial control signals, priority weight 0.8), video streaming services (monitoring video, weight 0.5), and file transfer services (log backup, weight 0.2). Using a priority sorting tool, the services are sorted from highest to lowest weight as follows: industrial control signals → monitoring video → log backup. Prioritize configuring back-off parameters for industrial control signals: allocate bandwidth of 200Mbps, set QoS level EF (accelerated forwarding), enable data caching (cache capacity of 100MB), and complete the back-off preparation for real-time interactive services.

[0068] Furthermore, based on the physical link interruption latency index as a latency impact factor, the corresponding service priority is corrected, and a gradual back-switch strategy is adopted to gradually switch the service data stream from the 5G network back to the physical link network according to the corrected service priority from high to low. After each batch of service switching is completed, the continuity and integrity parameters of service transmission are detected. In this embodiment of the invention, the current physical link interruption latency index of 0.15 (range 0-1, the lower the index, the more stable the link) is obtained, and 0.85 is calculated according to the formula "latency impact factor = 1 - interruption latency index". This factor is coupled with the service priority weight, and the corrected priorities are: industrial control signal = 0.8 × 0.85 = 0.68, monitoring video = 0.5 × 0.85 = 0.425, log backup = 0.2 × 0.85 = 0.17, while keeping the original order unchanged. A gradual switching strategy is adopted, switching 20% ​​of the service traffic in each batch: the first batch (1-2 minutes) switches 20% of the industrial control signal traffic (40Mbps), forwarding it from the 5G network (5G module IP192.168.2.100) to the physical link (GE0 / 0 interface IP192.168.1.100); after the switch is completed, the continuity (no disconnection, transmission latency reduced from 50ms to 15ms) and integrity (continuous sequence number of received data packets, no loss) are detected by transmission detection tools. The second batch (minutes 3-4) switched the remaining 80% traffic of the industrial control signal + 20% traffic of the monitoring video (160+80=240Mbps), and the test parameters were normal; the third batch (minutes 5-6) switched the remaining 80% traffic of the monitoring video + 20% traffic of the log backup (320+20=340Mbps); the fourth batch (minutes 7-8) switched the remaining 80% traffic of the log backup (80Mbps). After each batch of switching, a continuity and integrity test was performed to ensure no service interruption.

[0069] Furthermore, once all service data streams have successfully switched back to the physical link network and the transmission status is stable, the data transmission channel of the 5G network is shut down. The total duration of this back-switch process, the number of service interruptions, and the amount of data loss are recorded. These parameters are then merged with the switching process log to optimize the threshold parameters and weight coefficients in the millisecond-level routing switching mechanism.

[0070] In this embodiment of the invention, after all service data streams (industrial control 200Mbps, surveillance video 400Mbps, log backup 100Mbps, totaling 700Mbps) are switched back to the physical link, continuous monitoring is performed for 10 minutes (minutes 9-18). The transmission status is stable (latency 12-15ms, packet loss rate 0%, bandwidth utilization 70% ≤ 1000Mbps). The 5G network data transmission channel is closed using the channel management tool: the connection between the 5G module and the core network is disconnected (sending a "detach" signaling message), the IP forwarding function of the 5G module is disabled, and the channel status is marked as "closed". The back-switch process parameters are recorded: total duration 18 minutes (10 minutes preparation + 8 minutes switching), 0 service interruptions, and 0 bytes of data loss. These parameters are merged with the switching process log (including switching time, traffic, and detection results for each batch) to generate a back-switch analysis report. Based on the report, the millisecond-level routing switching mechanism has been optimized: the original back-switch trigger threshold of 0.87 has been adjusted to 0.85 (because back-switch is more stable during this low-traffic period), and the proportion of real-time interaction in the business priority weight has been increased from 0.8 to 0.85 to ensure that subsequent back-switch is more adapted to business needs.

[0071] Furthermore, the present invention also provides a network routing intelligent handover system based on 5G network and physical link, used to execute the network routing intelligent handover method based on 5G network and physical link as described above. The network routing intelligent handover system based on 5G network and physical link includes: The routing device data processing module is used to deploy routing devices with dual exits to access the physical link network and the 5G network respectively. It collects link status parameters of the physical link network, wireless transmission parameters of the 5G network, and service data flow characteristic parameters flowing through the routing device in real time through the routing device. It performs standardized preprocessing on the collected link status parameters, wireless transmission parameters, and service data flow characteristic parameters to generate preprocessed physical link status datasets, 5G network transmission datasets, and service characteristic datasets. The routing handover trigger evaluation module is used to obtain the packet loss rate trend characteristics, latency jitter characteristics, and bandwidth utilization characteristics of the physical link network based on the physical link status dataset, and calculate the physical link interruption lag index; extract the signal strength stability characteristics, interference level characteristics, and transmission rate fluctuation characteristics of the 5G network based on the 5G network transmission dataset, and evaluate and generate the 5G network availability order coefficient; construct the network routing handover trigger matrix based on the physical link interruption lag index and the 5G network availability order coefficient, and calculate the routing handover trigger threshold. The network routing link switching module is used to monitor the comparison between the physical link interruption latency index and the routing switching trigger threshold in real time. When the physical link interruption latency index is lower than the routing switching trigger threshold, a millisecond-level routing switching mechanism is activated to seamlessly switch the service data flow corresponding to the routing device from the physical link network to the 5G network. The network routing back-off response module is used to continuously monitor the status recovery parameters of the physical link network after the service data flow is switched to the 5G network and generate a physical link stability verification report through multiple rounds of bidirectional transmission tests. Based on the physical link stability verification report, it determines whether the physical link network has reached a stable operating state. If it has, it triggers the network routing back-off mechanism to smoothly switch the service data flow corresponding to the routing device from the 5G network back to the physical link network.

[0072] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features of the invention herein.

Claims

1. A method for intelligent network routing switching based on 5G network and physical link, characterized in that, Includes the following steps: Step S1: Deploy a routing device with dual exits to access the physical link network and the 5G network respectively. The routing device collects the link status parameters of the physical link network, the wireless transmission parameters of the 5G network, and the service data flow characteristic parameters flowing through the routing device in real time. The collected link status parameters, wireless transmission parameters, and service data flow characteristic parameters are standardized and preprocessed to generate preprocessed physical link status datasets, 5G network transmission datasets, and service characteristic datasets. Step S2: Based on the physical link status dataset, obtain the packet loss rate trend characteristics, latency jitter characteristics, and bandwidth utilization characteristics of the physical link network, and calculate the physical link interruption latency index; Based on the 5G network transmission dataset, the signal strength stability characteristics, interference level characteristics, and transmission rate fluctuation characteristics of the 5G network are extracted, and the availability order coefficient of the 5G network is evaluated and generated. Based on the physical link interruption delay index and the availability order coefficient of the 5G network, a network routing handover trigger matrix is ​​constructed, and the routing handover trigger threshold is calculated. Step S3: Monitor the comparison between the physical link interruption latency index and the route switching trigger threshold in real time. When the physical link interruption latency index is lower than the route switching trigger threshold, start the millisecond-level route switching mechanism to seamlessly switch the service data flow corresponding to the routing device from the physical link network to the 5G network. Step S4: After the service data stream is switched to the 5G network, continuously monitor the status recovery parameters of the physical link network and generate a physical link stability verification report through multiple rounds of bidirectional transmission tests; based on the physical link stability verification report, determine whether the physical link network has reached a stable operating state. If it has, trigger the network route back-switching mechanism to smoothly switch the service data stream corresponding to the routing device from the 5G network back to the physical link network.

2. The intelligent network routing switching method based on 5G network and physical link according to claim 1, characterized in that, Step S1 includes the following steps: Step S11: Establish connections with the physical link network and the 5G network respectively through the dual exit interfaces of the routing device, and configure the data acquisition module to periodically collect the link layer protocol status, frame check error rate and link negotiation status parameters of the physical link network in real time, so as to generate the link status parameters of the physical link network. Step S12: Real-time acquisition of reference signal received power, signal-to-interference-plus-noise ratio, cell handover frequency and network attachment status parameters corresponding to the 5G network through the 5G module to generate wireless transmission parameters of the 5G network. Step S13: Extract the protocol type, packet size distribution, transmission rate, and quality of service requirements parameters of the service data stream flowing through the routing device using deep packet inspection technology, in order to generate the characteristic parameters of the service data stream flowing through the routing device; Step S14: Perform outlier filtering on the collected link status parameters, wireless transmission parameters, and service data stream characteristic parameters to remove pulse-like abnormal data caused by transient interference, and obtain various parameters after outlier filtering. Step S15: Perform time axis alignment processing on various parameters after outlier filtering to unify the data sampling frequency, and transform parameters of different dimensions to the same data range through feature normalization to generate preprocessed physical link status dataset, 5G network transmission dataset and service feature dataset.

3. The intelligent network routing switching method based on 5G network and physical link according to claim 1, characterized in that, Step S2 includes the following steps: Step S21: Extract the data packet transmission records corresponding to the continuous time series from the physical link status dataset, and identify the timestamps and sequence number distributions of lost data packets by comparing the sent data packet sequence and the received data packet sequence; calculate the mean of the packet loss rate sliding window based on the timestamps and sequence number distributions of lost data packets and the clustering density on the time axis, and generate a linear trend coefficient of the packet loss rate as a trend feature of the packet loss rate through trend fitting. Step S22: Extract the transmission delay data of each data packet from the physical link state dataset and calculate the delay difference sequence corresponding to adjacent data packets; generate the delay dispersion parameter by statistically analyzing the standard deviation and coefficient of variation of the delay difference sequence; identify the periodic jitter component by combining the spectrum analysis of the transmission delay data, and use the amplitude ratio between the delay dispersion parameter and the periodic jitter component as the delay jitter feature; Step S23: Calculate the bandwidth occupancy ratio per unit time based on the total bandwidth capacity parameter and real-time transmission rate data corresponding to the physical link network, and extract the peak, valley and average values ​​corresponding to the bandwidth occupancy ratio through a sliding window. At the same time, generate the bandwidth saturation frequency parameter by combining the duration distribution corresponding to the bandwidth occupancy ratio. Use the weighted result of the peak-to-mean ratio, valley-to-mean ratio and bandwidth saturation frequency parameter corresponding to the bandwidth occupancy ratio as the bandwidth utilization feature. Step S24: Calculate the physical link interruption latency index based on the packet loss rate trend characteristics, latency jitter characteristics, and bandwidth utilization characteristics of the physical link network; Step S25: Extract the signal strength stability features, interference level features, and transmission rate fluctuation features of the 5G network based on the 5G network transmission dataset, and evaluate and generate the 5G network availability order coefficient; Step S26: Construct a network routing handover trigger matrix based on the physical link interruption delay index and the 5G network availability order coefficient, and calculate the routing handover trigger threshold.

4. The intelligent network routing switching method based on 5G network and physical link according to claim 3, characterized in that, Step S24 includes the following steps: Step S241: Perform time gradient analysis on the packet loss rate trend characteristics to generate a packet loss deterioration rate parameter by calculating the change in the linear trend coefficient of the packet loss rate per unit time. Combine the critical correlation threshold between packet loss rate and physical link interruption in historical data to map the packet loss deterioration rate parameter to the interruption warning coefficient corresponding to the packet loss dimension. Step S242: Generate a delay jitter frequency based on the delay dispersion parameter and periodic jitter component in the delay jitter characteristics, and calculate the degree of deviation between the delay jitter frequency and the preset service tolerance threshold to generate a delay risk index. At the same time, couple the delay risk index with the retransmission mechanism response speed of the physical link protocol to obtain the interruption warning coefficient corresponding to the delay dimension. Step S243: Analyze the peak-to-mean ratio, valley-to-mean ratio and bandwidth saturation frequency parameters in the bandwidth utilization characteristics to obtain the continuous growth trend corresponding to the bandwidth saturation state, and generate the bandwidth depletion acceleration parameter based on the continuous growth trend corresponding to the bandwidth saturation state. Combine the dynamic adjustment capability parameter corresponding to the physical link bandwidth to obtain the interruption warning coefficient corresponding to the bandwidth dimension. Step S244: The interruption warning coefficients corresponding to packet loss, latency and bandwidth dimensions are fused using the improved DS evidence theory. The weight ratio of each dimension warning coefficient is corrected by introducing a feature conflict resolution factor to generate a physical link comprehensive degradation index. Step S245: Based on the physical link comprehensive degradation index and combined with the time distribution of physical links from the same degree of degradation to complete interruption in historical data, the interruption delay assessment calculation is performed to obtain the physical link interruption delay index.

5. The intelligent network routing switching method based on 5G network and physical link according to claim 4, characterized in that, Step S245 includes the following steps: By extracting all complete link degradation-disruption event sequences from the historical operation logs of physical links, each sequence contains feature parameters of the entire process from the starting point of the same degree of degradation to complete interruption, and classifying these sequences according to the evolution curve shape of the physical link comprehensive degradation index, several typical degradation modes are generated, with each mode corresponding to a set of feature parameter distribution patterns. The time series of the comprehensive degradation index corresponding to the current physical link is matched with various typical degradation modes to calculate the dynamic time warping distance between the current sequence and each typical degradation mode. The typical degradation mode with the smallest distance is selected as the benchmark reference mode. At the same time, the subsequent time evolution parameters of the nodes with the same degradation degree as the current nodes in the benchmark reference mode are extracted. Based on the historical data statistics corresponding to the benchmark reference mode, the time sample set from nodes with the same degree of degradation to complete interruption is statistically analyzed, and the time distribution probability density function is generated by kernel density estimation. At the same time, the expected time value and confidence interval parameter corresponding to the function are calculated to obtain the historical average time delay benchmark value. By introducing a real-time environmental correction factor corresponding to the current physical link, which is calculated based on the temperature drift parameter, electromagnetic interference intensity and link load fluctuation characteristics of the current physical link, and converting the real-time environmental correction factor into a time adjustment coefficient through nonlinear mapping, the historical average time delay benchmark value is dynamically corrected based on the time adjustment coefficient to generate the corrected time delay benchmark value. The instantaneous rate of change of the current comprehensive degradation index is generated by using the subsequent time evolution parameters of nodes with the same degree of current degradation. The acceleration parameters corresponding to the link degradation process are calculated by combining the corrected time delay benchmark value. At the same time, the acceleration parameters are coupled with the corrected time delay benchmark value to obtain the predicted remaining time of the physical link from the current state to complete interruption. After normalization, the physical link interruption time delay index is generated.

6. The intelligent network routing switching method based on 5G network and physical link according to claim 3, characterized in that, Step S25 includes the following steps: Extract continuous time series reference signal received power data from 5G network transmission datasets, calculate the mean drift and variance rate of change of signal strength, combine signal quality mutation point detection to identify step change events of signal strength, and weight and fuse the mean drift, variance rate of change and the occurrence frequency of step change events to generate signal strength stability features. Based on the signal-to-interference-plus-noise ratio in the 5G network transmission data set, the spectral distribution characteristics of the interference signal are separated, and the instantaneous frequency and energy ratio of the interference signal are extracted by wavelet transform. The distribution entropy value of the interference signal in different frequency bands is calculated. Combined with the interference duration and the overlap parameter of the service frequency band, the interference level assessment matrix is ​​constructed, and the eigenvalues ​​of the interference level assessment matrix are used as the interference level features. Real-time sampling sequences of transmission rate are extracted from 5G network transmission datasets. Rate fluctuation sequences are generated by calculating the rate change of adjacent sampling points. The rate fluctuation sequences are decomposed into several intrinsic mode components. The periodic characteristics and energy proportion of each component are analyzed. The dominant fluctuation component is selected and its amplitude variation coefficient is calculated. The amplitude variation coefficient is coupled with the Hurst exponent of the rate fluctuation to generate transmission rate fluctuation characteristics. A three-dimensional space for 5G network availability assessment is constructed, with signal strength stability characteristics, interference level characteristics, and transmission rate fluctuation characteristics as coordinate axes. Availability benchmark points in the space are calibrated using historical communication quality data. The Mahalanobis distance between the current network status and the benchmark point in the three-dimensional space is calculated, and this distance is converted into an initial availability score. A network load resilience coefficient is introduced to correct the initial availability score. This coefficient is calculated based on the current number of user connections, cell handover success rate and resource block utilization of the 5G network. By establishing a mapping relationship between the network load resilience coefficient and the availability decay factor, the initial availability score is dynamically adjusted to generate the 5G network availability order coefficient.

7. The intelligent network routing switching method based on 5G network and physical link according to claim 3, characterized in that, Step S26 includes the following steps: Extract the exponential change slope parameter from the time series of the physical link interruption delay index, and generate the link deterioration urgency coefficient by analyzing the absolute value growth trend of the exponential change slope parameter. At the same time, based on the historical fluctuation range of the 5G network availability order coefficient, calculate the deviation parameter between the current availability order coefficient and the optimal value to obtain the 5G network service capacity margin. A two-dimensional feature mapping space is constructed, with the link deterioration urgency coefficient as the vertical axis and the 5G network service capacity margin as the horizontal axis. The historical handover decision data is mapped into several decision-dense regions in the space through the K-means clustering algorithm. Each dense region corresponds to a set of typical handover strategy labels, generating a decision reference map. Based on the boundary parameters of each dense area in the decision reference map, a network routing handover trigger matrix is ​​established. The row dimension of the matrix corresponds to the graded interval of the urgency of link deterioration, the column dimension corresponds to the graded interval of the 5G network service capacity margin, and the matrix element is the coupling result of the historical handover success rate and the service interruption risk value under the combination of the interval. A transmission demand sensitivity parameter corresponding to the business characteristic dataset is introduced. This parameter is calculated based on the real-time weight of the business, data integrity requirements and bandwidth dependence. The transmission demand sensitivity parameter is converted into a matrix correction factor through fuzzy hierarchical analysis to dynamically adjust the element values ​​in the network routing switching trigger matrix. A multi-objective optimization algorithm is used to solve the Pareto optimal solution of the adjusted matrix. The weighted product of the link deterioration urgency coefficient and the 5G network service capacity margin is used as the objective function. Combined with the constraint of service interruption risk, the critical decision thresholds under different business scenarios are calculated. These critical decision thresholds are normalized and fused to generate the routing handover trigger threshold.

8. The intelligent network routing switching method based on 5G network and physical link according to claim 1, characterized in that, Step S4 includes the following steps: Step S41: After the service data stream is switched to the 5G network, continuously monitor and collect the status recovery parameters corresponding to the physical link network, including link connectivity parameters, data forwarding success rate and protocol negotiation status parameters; Step S42: Determine whether the physical link network has recovered from the fault state based on the collected state recovery parameters. When the link connectivity parameters meet the preset conditions and the data forwarding success rate reaches the recovery threshold, it is determined that the physical link network has entered the state recovery period. Step S43: During the state recovery period, start the link stability verification process to conduct multiple rounds of bidirectional transmission tests by sending test data packets of different sizes and protocol types, and record the packet loss rate, latency and jitter parameters of each round of test; Step S44: Calculate the stability score of the physical link network based on the packet loss rate, latency and jitter parameters of multiple rounds of testing. When the stability score of multiple consecutive rounds of testing is higher than the stability threshold, generate a physical link stability verification report. The physical link stability verification report includes the link recovery time, stability and recommended back-off timing parameters. Step S45: Based on the physical link stability verification report, determine whether the physical link network has reached a stable operating state. If it has, trigger the network route back-switching mechanism to smoothly switch the service data flow corresponding to the routing device from the 5G network back to the physical link network.

9. The intelligent network routing switching method based on 5G network and physical link according to claim 8, characterized in that, Step S45 includes the following steps: The system analyzes the parameters in the physical link stability verification report. When the stability of the physical link network reaches the preset back-off standard and the recommended back-off timing parameter indicates that the current time is the best back-off window, the back-off preparation process is started. In the back-switching preparation process, a data transmission channel corresponding to the physical link network is pre-established, and the service data streams on the 5G network are sorted based on the service priority parameters in the service feature dataset, with priority given to preparing back-switching for real-time interactive services. The service priority is corrected based on the physical link interruption time delay index as the time delay impact factor. A gradual back-switch strategy is adopted to switch the service data stream from the 5G network back to the physical link network in order of the corrected service priority from high to low. After each batch of service switching is completed, the continuity and integrity parameters of service transmission are detected. Once all service data streams have successfully switched back to the physical link network and the transmission status is stable, the data transmission channel of the 5G network is shut down. The total duration of this switchback process, the number of service interruptions, and the amount of data loss are recorded. These parameters are then merged with the switchback process log to optimize the threshold parameters and weight coefficients in the millisecond-level routing switchback mechanism.

10. A network routing intelligent switching system based on 5G network and physical link, characterized in that, For executing the intelligent network routing handover method based on 5G network and physical link as described in claim 1, the intelligent network routing handover system based on 5G network and physical link includes: The routing device data processing module is used to deploy routing devices with dual exits to access the physical link network and the 5G network respectively. It collects link status parameters of the physical link network, wireless transmission parameters of the 5G network, and service data flow characteristic parameters flowing through the routing device in real time through the routing device. It performs standardized preprocessing on the collected link status parameters, wireless transmission parameters, and service data flow characteristic parameters to generate preprocessed physical link status datasets, 5G network transmission datasets, and service characteristic datasets. The routing handover trigger evaluation module is used to obtain the packet loss rate trend characteristics, latency jitter characteristics, and bandwidth utilization characteristics of the physical link network based on the physical link status dataset, and calculate the physical link interruption lag index; extract the signal strength stability characteristics, interference level characteristics, and transmission rate fluctuation characteristics of the 5G network based on the 5G network transmission dataset, and evaluate and generate the 5G network availability order coefficient; construct the network routing handover trigger matrix based on the physical link interruption lag index and the 5G network availability order coefficient, and calculate the routing handover trigger threshold. The network routing link switching module is used to monitor the comparison between the physical link interruption latency index and the routing switching trigger threshold in real time. When the physical link interruption latency index is lower than the routing switching trigger threshold, a millisecond-level routing switching mechanism is activated to seamlessly switch the service data flow corresponding to the routing device from the physical link network to the 5G network. The network routing back-off response module is used to continuously monitor the status recovery parameters of the physical link network after the service data flow is switched to the 5G network and generate a physical link stability verification report through multiple rounds of bidirectional transmission tests. Based on the physical link stability verification report, it determines whether the physical link network has reached a stable operating state. If it has, it triggers the network routing back-off mechanism to smoothly switch the service data flow corresponding to the routing device from the 5G network back to the physical link network.

Images