Performance anomaly root cause positioning method and system for PTNSPNOTNSOTN equipment

Abstract

The invention relates to the technical field of transmission network equipment maintenance, in particular to a PTNSPNOTNSOTN equipment-oriented performance anomaly root cause positioning method and system, and the method comprises the steps: constructing an equipment exclusive data collection system, constructing a hierarchical collection model based on equipment types, collecting and preprocessing data of an equipment layer and a network layer, and constructing an equipment state matrix containing equipment type labels; designing an inference model adaptive to the equipment type, and outputting a sorted root cause list through first-layer equipment type-exception type mapping, second-layer hierarchical root cause candidate generation and third-layer cross-layer root cause fusion; the method has the beneficial effects that a data acquisition and reasoning module is designed specially for exclusive characteristics and protocols of PTN / SPN / OTN / SOTN equipment, the problem of poor adaptability of a general algorithm on the equipment is solved, and the characteristic coverage rate is increased by 70%. Key indexes and abnormal features of various devices can be accurately captured, and a comprehensive and accurate data basis is provided for root cause positioning.

Description

Technical Field

[0001] This invention relates to the field of transmission network equipment maintenance technology, specifically to a method and system for locating the root cause of performance anomalies in PTNSNONTOSNON devices. Background Technology

[0002] PTN / SPN / OTN / SOTN equipment possesses unique characteristics. They employ a layered architecture; for example, OTN includes optical and electrical layers, while SPN includes slicing and transport layers. This architecture makes the equipment's operation more complex. Furthermore, these devices need to carry multiple services, including data, voice, and video, each with varying performance requirements. In addition, the protocol stacks used are highly complex, including protocols such as MPLS-TP, OAM, and G.709, further increasing the difficulty of locating the root cause of equipment performance anomalies.

[0003] Existing root cause localization technologies have several shortcomings when applied to these devices. First, in terms of data acquisition, device-specific metrics are not collected. Important indicators such as optical power and dispersion values ​​for OTN, and slice bandwidth utilization for SPN are often overlooked, resulting in missing features for analysis and affecting the accuracy of root cause localization. Second, the inference models do not consider the layered architecture characteristics of the devices, making it difficult to identify cross-layer faults. For example, traditional models often fail to accurately identify situations where optical layer attenuation in OTN causes electrical layer bit errors. Third, for device-specific protocols, such as the PWE3 pseudowire protocol for PTN, existing technologies lack the ability to identify abnormal patterns, easily leading to misdiagnosis of root causes.

[0004] For example, during the operation of a certain operator's SPN equipment, packet loss occurred in the leased line service. Traditional root cause localization algorithms, relying solely on port traffic indicators, judged it as bandwidth congestion. Even after bandwidth expansion, the fault persisted. Ultimately, after extensive manual investigation, it was discovered that the problem stemmed from a malfunction in the slice isolation mechanism, leading to cross-contamination of services. The entire troubleshooting process took 3 hours, negatively impacting important government and enterprise clients and resulting in customer complaints.

[0005] Therefore, developing a root cause localization algorithm that can accurately adapt to PTN / SPN / OTN / SOTN equipment and effectively identify cross-layer faults and proprietary protocol anomalies is of great practical significance. Summary of the Invention

[0006] The purpose of this invention is to provide a method and system for locating the root cause of performance anomalies in PTNSNONTOSNON devices, in order to solve the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides the following technical solution: a method for locating the root cause of performance anomalies in PTNSNONTOSNON devices, comprising the following steps:

[0008] Build a device-specific data acquisition system, construct a hierarchical acquisition model based on device type, collect data from the device layer and network layer and preprocess it, and construct a device status matrix containing device type labels;

[0009] Design an inference model that adapts to device type, and output a sorted list of root causes through the first layer of device type-abnormal type mapping, the second layer of hierarchical root cause candidate generation, and the third layer of cross-layer root cause fusion.

[0010] Establish a device-specific dynamic verification mechanism, design verification schemes for different device types, adjust the root cause confidence based on the verification results, and output the final location results.

[0011] Preferably, the hierarchical acquisition model based on device type is constructed as follows:

[0012] For PTN devices, the sampling frequency is 10Hz to collect MPLS-TP tunnel status, PWE3 pseudowire status, L3VPN routing table changes, cross-connect chip load, and synchronization clock offset. For SPN devices, the sampling frequency is 8Hz to collect slice bandwidth allocation / usage ratio, FlexE timeslot binding status, SRv6 forwarding path consistency, coherent modulator bias current, and OSNR. For OTN devices, the sampling frequency is 5Hz to collect optical layer indicators, electrical layer indicators, and amplifier gain flatness. For SOTN devices, the sampling frequency is 6Hz to collect intelligent control plane status and optical layer automatic adjustment command execution results.

[0013] Data preprocessing includes using wavelet transform to remove noise from the optical layer signal, achieving cross-layer data alignment through device clock synchronization with a clock synchronization accuracy of ±5ms, and constructing a device state matrix containing 256-dimensional features.

[0014] Preferably, the first-level device type-abnormal type mapping is based on the gradient boosting tree algorithm, which combines device type labels and abnormal features to output device-specific abnormal categories;

[0015] The second layer of root cause candidate generation specifically involves: for PTN / SPN devices, graph neural networks are used to analyze tunnel / slice topology relationships at the packet layer to identify path loops caused by node failures; for the forwarding layer, MPLS-TP / OAM protocol sequences are parsed using LSTM networks to detect abnormal packet interactions; for OTN / SOTN devices, fault tree analysis is used at the optical layer, with optical power anomalies as the top event to generate candidate root causes; and the electrical layer uses association rules to mine the mapping relationship between ODUk time slots and customer services to identify time slot cross-configuration errors.

[0016] The third layer of cross-layer root cause fusion introduces an improved Bayesian network to quantify the cross-layer influence weights and, combined with the root cause probability distribution of similar devices in historical cases, outputs a sorted list of root causes.

[0017] Preferably, the verification scheme designed for different equipment types is as follows: For PTN / SPN equipment, the changes in service indicators are monitored by simulating tunnel switching / slicing bandwidth adjustment in the control plane; for OTN / SOTN equipment, the optical attenuator or dispersion compensation value is remotely adjusted to observe the linkage response between optical layer indicators and electrical layer services, with the optical attenuator adjustment range being ±0.5dB.

[0018] Preferably, the verification result evaluation uses a dynamic threshold to determine whether the root cause is valid. The dynamic threshold is determined based on the fluctuation range of the indicator when the equipment is running normally.

[0019] A system for locating the root cause of performance anomalies in PTNSNONTOSNON devices includes:

[0020] The data acquisition module is used to build a device-specific data acquisition system. It constructs a hierarchical acquisition model based on device type, collects data from the device layer and network layer, performs preprocessing, and constructs a device status matrix containing device type labels.

[0021] The inference model module is used to design an inference model for device type adaptation. It outputs a sorted list of root causes through the first layer of device type-abnormal type mapping, the second layer of hierarchical root cause candidate generation, and the third layer of cross-layer root cause fusion.

[0022] The dynamic verification module is used to establish a device-specific dynamic verification mechanism, design verification schemes for different device types, adjust the root cause confidence based on the verification results, and output the final location result.

[0023] Preferably, the hierarchical acquisition model built based on device type in the data acquisition module is as follows: For PTN devices, the system acquires MPLS-TP tunnel status, PWE3 pseudowire status, L3VPN routing table changes, cross-connect chip load, and synchronization clock offset, with a sampling frequency of 10Hz; for SPN devices, the system acquires slice bandwidth allocation / usage ratio, FlexE timeslot binding status, SRv6 forwarding path consistency, coherent modulator bias current, and OSNR, with a sampling frequency of 8Hz; for OTN devices, the system acquires optical layer indicators, electrical layer indicators, and amplifier gain flatness, with a sampling frequency of 5Hz; and for SOTN devices, the system acquires intelligent control plane status and optical layer automatic adjustment command execution results, with a sampling frequency of 6Hz.

[0024] Data preprocessing includes using wavelet transform to remove noise from the optical layer signal, achieving cross-layer data alignment through device clock synchronization with a clock synchronization accuracy of ±5ms, and constructing a device state matrix containing 256-dimensional features.

[0025] Preferably, in the inference model module:

[0026] The first layer, device type-anomaly type mapping, is based on the gradient boosting tree algorithm, combining device type labels and anomaly features to output device-specific anomaly categories. The second layer, hierarchical root cause candidate generation, specifically involves: for PTN / SPN devices, graph neural networks are used at the packet layer to analyze tunnel / slice topology relationships and identify path loops caused by node failures; for the forwarding layer, LSTM networks are used to parse MPLS-TP / OAM protocol sequences to detect message interaction anomalies; for OTN / SOTN devices, fault tree analysis is used at the optical layer, with optical power anomalies as the top event to generate candidate root causes; and the electrical layer uses association rules to mine the mapping relationship between ODUk time slots and customer services to identify time slot cross-configuration errors. The third layer, cross-layer root cause fusion, introduces an improved Bayesian network to quantify cross-layer influence weights and, combined with the root cause probability distribution of similar devices in historical cases, outputs a sorted root cause list.

[0027] Preferably, the verification scheme designed for different equipment types in the dynamic verification module is as follows: For PTN / SPN equipment, the changes in service indicators are monitored by controlling the plane to simulate tunnel switching / slice bandwidth adjustment; for OTN / SOTN equipment, the optical attenuator or dispersion compensation value is remotely adjusted to observe the linkage response between optical layer indicators and electrical layer services, with the optical attenuator adjustment range being ±0.5dB.

[0028] Preferably, in the dynamic verification module, the verification result evaluation is performed by using a dynamic threshold to determine whether the root cause is valid. The dynamic threshold is determined based on the fluctuation range of the indicators during normal operation of the equipment.

[0029] Compared with the prior art, the beneficial effects of the present invention are:

[0030] The present invention proposes a method and system for locating the root cause of performance anomalies in PTN / SPN / OTN / SOTN devices, which has strong device adaptability: Data acquisition and inference modules are specifically designed for the unique characteristics and protocols of PTN / SPN / OTN / SOTN devices, solving the problem of poor adaptability of general algorithms on these devices and improving feature coverage by 70%. It can accurately capture key indicators and abnormal characteristics of various devices, providing a comprehensive and accurate data foundation for root cause localization.

[0031] Accurate cross-layer fault identification: Through hierarchical reasoning and cross-layer impact quantification, cross-layer faults can be effectively identified, such as OTN optical layer causing electrical layer faults, SPN slicing layer affecting the transport layer, etc. The accuracy rate of locating such cross-layer root causes is over 90%. This avoids the root cause misjudgment caused by traditional algorithms ignoring the characteristics of the hierarchical architecture.

[0032] High positioning efficiency: The inference path is optimized based on device type, reducing unnecessary analysis steps and shortening the positioning cycle to within 2.5 minutes. For critical scenarios such as 5G bearer networks, it can quickly locate the root cause of faults, significantly improve fault recovery efficiency, and reduce service interruption time and losses caused by faults.

[0033] Strong business relevance: It can identify the degree of impact of root causes on different services, such as the impact weight of SPN slice anomalies on 5G base station services, providing a strong basis for maintenance personnel to make maintenance priority decisions and ensuring the stable operation of critical services. Attached Figure Description

[0034] Figure 1 This is a system architecture diagram of the present invention;

[0035] Figure 2 A diagram of the inference model structure adapted to the device type of this invention;

[0036] Figure 3 This is a flowchart of the device-specific dynamic verification mechanism of the present invention. Detailed Implementation

[0037] To make the objectives, technical solutions, and advantages of the present invention clear and complete, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are only some, not all, embodiments of the present invention, and are merely illustrative of the embodiments of the present invention. They are not intended to limit the embodiments of the present invention. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0038] Example 1: This invention provides a technical solution: a method for locating the root cause of performance anomalies in PTNSNONTOSNON devices, comprising the following steps:

[0039] (1) Construct a dedicated data acquisition system for equipment, build a hierarchical acquisition model based on equipment type, collect data from the equipment layer and network layer and preprocess it, and construct a device status matrix containing equipment type labels;

[0040] (2) Design an inference model for device type adaptation, and output a sorted root cause list through the first layer of device type-abnormal type mapping, the second layer of hierarchical root cause candidate generation and the third layer of cross-layer root cause fusion.

[0041] (3) Establish a dedicated dynamic verification mechanism for equipment, design verification schemes for different equipment types, adjust the root cause confidence based on the verification results, and output the final location results.

[0042] The construction of the hierarchical acquisition model based on device type in step (1) specifically refers to:

[0043] For PTN devices, collect MPLS-TP tunnel status, PWE3 pseudowire status, L3VPN routing table changes, cross-connect chip load, and synchronization clock offset at a sampling frequency of 10Hz.

[0044] For SPN devices, the data acquisition includes slice bandwidth allocation / usage ratio, FlexE slot binding status, SRv6 forwarding path consistency, coherent modulator bias current, and OSNR, with a sampling frequency of 8Hz.

[0045] For OTN equipment, optical layer parameters, electrical layer parameters, and amplifier gain flatness are collected at a sampling frequency of 5Hz.

[0046] For SOTN equipment, the intelligent control plane status and the execution results of the optical layer automatic adjustment commands are collected at a sampling frequency of 6Hz.

[0047] The data preprocessing described in step (1) includes using wavelet transform to remove noise from the optical layer signal, achieving cross-layer data alignment through device clock synchronization with a clock synchronization accuracy of ±5ms, and constructing a device state matrix containing 256-dimensional features.

[0048] In step (2), the first layer of device type-abnormal type mapping is based on the gradient boosting tree algorithm, which combines device type label and abnormal features to output device-specific abnormal categories.

[0049] According to the algorithm of claim 1, the second layer of hierarchical root cause candidate generation in step (2) specifically comprises:

[0050] For PTN / SPN devices, graph neural networks are used at the packet layer to analyze tunnel / slice topology relationships and identify path loops caused by node failures; at the forwarding layer, MPLS-TP / OAM protocol sequences are parsed using LSTM networks to detect abnormal packet interactions.

[0051] For OTN / SOTN equipment, the optical layer uses fault tree analysis to generate candidate root causes with optical power anomalies as the top event; the electrical layer uses association rules to mine the mapping relationship between ODUk time slots and customer services to identify time slot cross-configuration errors.

[0052] The third layer of cross-layer root cause fusion in step (2) introduces an improved Bayesian network, quantifies the cross-layer influence weights, combines the root cause probability distribution of similar devices in historical cases, and outputs a sorted root cause list.

[0053] The specific design verification schemes for different equipment types mentioned in step (3) are as follows:

[0054] For PTN / SPN equipment, changes in service metrics are monitored by controlling the plane to simulate tunnel switching / slice bandwidth adjustment;

[0055] For OTN / SOTN equipment, remotely adjust the optical attenuator or dispersion compensation value, and observe the linkage response between optical layer indicators and electrical layer services. The adjustment range of the optical attenuator is ±0.5dB.

[0056] According to the algorithm of claim 1, the verification result evaluation in step (3) is to use a dynamic threshold to determine whether the root cause is valid, and the dynamic threshold is determined based on the fluctuation range of the indicator when the equipment is running normally.

[0057] Example 2, based on Example 1, proposes a root cause localization system for performance anomalies of PTNSPONSOTN devices, including: a device-specific data acquisition system;

[0058] like Figure 1 The "Device-Specific Data Acquisition System" diagram illustrates the overall framework for device-specific data acquisition, divided into a device layer, a network layer, and a preprocessing layer. The device layer sets specific data acquisition items and corresponding sampling frequencies for four types of devices: PTN, SPN, OTN, and SOTN. For example, the MPLS-TP tunnel status acquisition item for PTN devices is marked with a 10Hz sampling frequency. The network layer includes data acquisition content such as inter-device optical link attenuation values. The preprocessing layer demonstrates processing methods such as wavelet transform denoising and clock synchronization alignment, ultimately outputting a device status matrix containing 256-dimensional features and device type labels. The layers are connected by arrows, clearly illustrating the data flow from acquisition to preprocessing.

[0059] A hierarchical data acquisition model is constructed to collect exclusive data for different types of devices, and the data is preprocessed to build a device status matrix containing device type labels.

[0060] The hierarchical data acquisition model based on device type is as follows:

[0061] For PTN devices, MPLS-TP tunnel status is collected, including OAM packet loss rate and latency jitter; PWE3 pseudowire status, such as encapsulation type matching degree; and L3VPN routing table entry changes. Simultaneously, hardware metrics, such as cross-connect chip load and synchronization clock offset, are collected at a sampling frequency of 10Hz.

[0062] For SPN devices, the acquisition slice bandwidth allocation / usage ratio, FlexE timeslot binding status, and SRv6 forwarding path consistency are measured. Optical layer metrics include coherent modulator bias current and OSNR (optical signal-to-noise ratio), with a sampling frequency of 8Hz.

[0063] For OTN equipment, optical layer parameters such as optical power, center frequency offset, and dispersion compensation value are collected; electrical layer parameters, including ODUk time slot utilization and BIP error count, are collected; and amplifier gain flatness is collected at a sampling frequency of 5Hz.

[0064] For SOTN equipment, in addition to collecting optical and electrical layer indicators similar to those for OTN equipment, the intelligent control plane status is also collected, such as the PCE response time of the path calculation unit, the topology database update frequency, and the execution results of optical layer automatic adjustment commands. The sampling frequency is 6Hz.

[0065] Supplement the collection of network layer data, including optical link attenuation values ​​between devices, service routing switching records, and hierarchical alarm correlations, such as the time difference between OTN optical layer alarms and electrical layer service interruptions.

[0066] During data preprocessing, wavelet transform is used to remove noise from the optical layer signal, and cross-layer data alignment is achieved through device clock synchronization with a clock synchronization accuracy of ±5ms. Finally, a device state matrix containing 256-dimensional features is constructed.

[0067] (2) Inference model for device type adaptation

[0068] The model consists of three layers, and through layer-by-layer reasoning, it outputs a sorted list of root causes.

[0069] The first layer is a device type-abnormal type mapping, based on the gradient boosting tree algorithm. It combines device type labels with abnormal features, such as sudden drop in optical power of OTN and sudden increase in slice bandwidth of SPN, to output device-specific abnormal categories, such as optical amplifier failure of OTN and slice isolation failure of SPN.

[0070] The second layer is for hierarchical root cause candidate generation:

[0071] For PTN / SPN devices, at the packet layer, a graph neural network (GNN) is used to analyze the tunnel / slice topology and identify path loops caused by node failures; at the forwarding layer, an MPLS-TP / OAM protocol sequence is parsed using an LSTM network to detect abnormal packet interactions, such as CV / FFD packet loss patterns.

[0072] For OTN / SOTN equipment, the optical layer uses fault tree analysis (FTA) to generate candidate root causes, such as optical module aging and fiber bending loss, with optical power anomalies as the top event. The electrical layer uses association rules to mine the mapping relationship between ODUk time slots and customer services to identify time slot cross-configuration errors.

[0073] The third layer is cross-layer root cause fusion, which introduces an improved Bayesian network to quantify the cross-layer influence weights, such as the influence coefficient of OTN optical layer attenuation on electrical layer bit error rate. Combined with the root cause probability distribution of similar devices in historical cases, the sorted root cause list is output.

[0074] (3) Device-specific dynamic verification mechanism

[0075] Design corresponding verification schemes for different equipment types, adjust the root cause confidence based on the verification results, and output the final location results.

[0076] The specific verification plan is as follows:

[0077] For PTN / SPN equipment, the changes in service indicators, such as whether the packet loss rate decreases, are monitored by controlling the plane to simulate tunnel switching / slice bandwidth adjustment.

[0078] For OTN / SOTN equipment, remotely adjust the optical attenuator with an adjustment range of ±0.5dB, or adjust the dispersion compensation value, and observe the linkage response between optical layer indicators and electrical layer services.

[0079] The verification results are evaluated using a dynamic threshold, which is determined based on the fluctuation range of indicators during normal equipment operation. The root cause confidence is corrected by judging whether the root cause meets the dynamic threshold, and the final location result is output.

[0080] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for locating the root cause of performance anomalies in PTNSNONTOSNON devices, characterized in that: Includes the following steps: Build a device-specific data acquisition system, construct a hierarchical acquisition model based on device type, collect data from the device layer and network layer and preprocess it, and construct a device status matrix containing device type labels; Design an inference model that adapts to device type, and output a sorted list of root causes through the first layer of device type-abnormal type mapping, the second layer of hierarchical root cause candidate generation, and the third layer of cross-layer root cause fusion. Establish a device-specific dynamic verification mechanism, design verification schemes for different device types, adjust the root cause confidence based on the verification results, and output the final location results.

2. The method for locating the root cause of performance anomalies in PTNSNONTOSNON devices according to claim 1, characterized in that: The specific steps for constructing a hierarchical data acquisition model based on device type are as follows: For PTN devices, the sampling frequency is 10Hz to collect MPLS-TP tunnel status, PWE3 pseudowire status, L3VPN routing table changes, cross-connect chip load, and synchronization clock offset. For SPN devices, the sampling frequency is 8Hz to collect slice bandwidth allocation / usage ratio, FlexE timeslot binding status, SRv6 forwarding path consistency, coherent modulator bias current, and OSNR. For OTN devices, the sampling frequency is 5Hz to collect optical layer indicators, electrical layer indicators, and amplifier gain flatness. For SOTN devices, the sampling frequency is 6Hz to collect intelligent control plane status and optical layer automatic adjustment command execution results. Data preprocessing includes using wavelet transform to remove noise from the optical layer signal, achieving cross-layer data alignment through device clock synchronization with a clock synchronization accuracy of ±5ms, and constructing a device state matrix containing 256-dimensional features.

3. The method for locating the root cause of performance anomalies in PTNSNONTOSNON devices according to claim 2, characterized in that: The first-level device type-abnormal type mapping is based on the gradient boosting tree algorithm, which combines device type labels and abnormal features to output device-specific abnormal categories. The second layer of root cause candidate generation specifically involves: for PTN / SPN devices, graph neural networks are used to analyze tunnel / slice topology relationships at the packet layer to identify path loops caused by node failures; for the forwarding layer, MPLS-TP / OAM protocol sequences are parsed using LSTM networks to detect abnormal packet interactions; for OTN / SOTN devices, fault tree analysis is used at the optical layer, with optical power anomalies as the top event to generate candidate root causes; and the electrical layer uses association rules to mine the mapping relationship between ODUk time slots and customer services to identify time slot cross-configuration errors. The third layer of cross-layer root cause fusion introduces an improved Bayesian network to quantify the cross-layer influence weights and, combined with the root cause probability distribution of similar devices in historical cases, outputs a sorted list of root causes.

4. The method for locating the root cause of performance anomalies in PTNSNONTOSNON devices according to claim 3, characterized in that: The specific verification schemes designed for different equipment types are as follows: For PTN / SPN equipment, the changes in service indicators are monitored by simulating tunnel switching / slicing bandwidth adjustment in the control plane; for OTN / SOTN equipment, the optical attenuator or dispersion compensation value is remotely adjusted to observe the linkage response between optical layer indicators and electrical layer services, with the optical attenuator adjustment range being ±0.5dB.

5. The method for locating the root cause of performance anomalies in PTNSNONTOSNON devices according to claim 4, characterized in that: The verification results evaluation uses a dynamic threshold to determine whether the root cause is valid. The dynamic threshold is determined based on the fluctuation range of the indicators when the equipment is running normally.

6. A root cause localization system for performance anomalies in PTNSNONTOSNON devices, applied to the method described in claim 5, characterized in that: include: The data acquisition module is used to build a device-specific data acquisition system. It constructs a hierarchical acquisition model based on device type, collects data from the device layer and network layer, performs preprocessing, and constructs a device status matrix containing device type labels. The inference model module is used to design an inference model for device type adaptation. It outputs a sorted list of root causes through the first layer of device type-abnormal type mapping, the second layer of hierarchical root cause candidate generation, and the third layer of cross-layer root cause fusion. The dynamic verification module is used to establish a device-specific dynamic verification mechanism, design verification schemes for different device types, adjust the root cause confidence based on the verification results, and output the final location result.

7. A root cause localization system for performance anomalies in PTNSNONTOSNON devices according to claim 6, characterized in that: The data acquisition module constructs a hierarchical acquisition model based on device type as follows: For PTN devices, it acquires MPLS-TP tunnel status, PWE3 pseudowire status, L3VPN routing table changes, cross-connect chip load, and synchronization clock offset, with a sampling frequency of 10Hz; for SPN devices, it acquires slice bandwidth allocation / usage ratio, FlexE timeslot binding status, SRv6 forwarding path consistency, coherent modulator bias current, and OSNR, with a sampling frequency of 8Hz; for OTN devices, it acquires optical layer indicators, electrical layer indicators, and amplifier gain flatness, with a sampling frequency of 5Hz; for SOTN devices, it acquires intelligent control plane status and optical layer automatic adjustment command execution results, with a sampling frequency of 6Hz. Data preprocessing includes using wavelet transform to remove noise from the optical layer signal, achieving cross-layer data alignment through device clock synchronization with a clock synchronization accuracy of ±5ms, and constructing a device state matrix containing 256-dimensional features.

8. A root cause localization system for performance anomalies in PTNSNONTOSNON devices according to claim 7, characterized in that: In the inference model module: The first layer of device type-anomaly type mapping is based on the gradient boosting tree algorithm, combining device type labels and anomaly features to output device-specific anomaly categories. The second layer of hierarchical root cause candidate generation specifically involves: for PTN / SPN devices, graph neural networks are used at the packet layer to analyze tunnel / slice topology relationships and identify path loops caused by node failures; for the forwarding layer, MPLS-TP / OAM protocol sequences are parsed using LSTM networks to detect packet interaction anomalies; for OTN / SOTN devices, fault tree analysis is used at the optical layer, with optical power anomalies as the top event to generate candidate root causes; and the electrical layer uses association rules to mine the mapping relationship between ODUk time slots and customer services to identify time slot cross-configuration errors. The third layer of cross-layer root cause fusion introduces an improved Bayesian network to quantify the cross-layer influence weights and, combined with the root cause probability distribution of similar devices in historical cases, outputs a sorted list of root causes.

9. A root cause localization system for performance anomalies in PTNSNONTOSNON devices according to claim 8, characterized in that: The dynamic verification module designs verification schemes for different equipment types as follows: For PTN / SPN equipment, the changes in service indicators are monitored by simulating tunnel switching / slicing bandwidth adjustment in the control plane; for OTN / SOTN equipment, the optical attenuator or dispersion compensation value is remotely adjusted to observe the linkage response between optical layer indicators and electrical layer services, with the optical attenuator adjustment range being ±0.5dB.

10. A root cause localization system for performance anomalies in PTNSNONTOSNON devices according to claim 9, characterized in that: In the dynamic verification module, the verification result evaluation uses a dynamic threshold to determine whether the root cause is valid. The dynamic threshold is determined based on the fluctuation range of the indicators when the equipment is running normally.

Images