Artificial intelligence-based network device failure prediction method, system and medium
By embedding judgment components in network devices to perform fault entropy determination and joint decision-making, the uncertainty problem of network device fault prediction is solved, achieving high-precision fault prediction and effective operation and maintenance management, thereby improving the stability and reliability of the devices.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SICHUAN CHANGFU INFORMATION TECHNOLOGY SERVICE CO LTD
- Filing Date
- 2025-10-16
- Publication Date
- 2026-05-19
AI Technical Summary
Existing network equipment fault prediction methods suffer from high uncertainty due to signal discreteness, resulting in low fault prediction accuracy, lagging maintenance and management, and a lack of foresight.
By collecting network device operation signals, using embedded judgment components to perform fault entropy determination, generating fault prediction instructions, and combining them with a predictive agent to make joint decisions based on physical laws and causal logic, a fault control scheme is determined.
It improves the accuracy of fault prediction and operational efficiency, reduces the frequency of equipment failures, and enhances the stability and reliability of network equipment.
Smart Images
Figure CN121333959B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of fault prediction technology, and in particular to a method, system and medium for predicting network device faults based on artificial intelligence. Background Technology
[0002] In today's era of rapid digital economic development, the stability and reliability of network infrastructure have become the cornerstone of societal operation. Various network devices, such as routers, switches, and servers, serve as critical nodes supporting massive data flows; any potential failure in these devices can trigger a chain reaction, leading to service interruptions, data loss, and even significant economic losses. Therefore, predicting network device failures and managing its health, shifting from reactive maintenance to proactive early warning, has become a core requirement in the field of network operations and maintenance.
[0003] Traditional fault prediction methods primarily rely on threshold-based monitoring or models based on historical statistical data. While threshold monitoring is simple and direct, it struggles to address the gradual and complex nature of equipment degradation, often triggering alarms only when fault symptoms are obvious, resulting in short warning windows and a lack of foresight. Early statistical models, on the other hand, are overly dependent on data quality and feature engineering, and struggle to capture the complex nonlinear relationships and hidden failure mechanisms during equipment operation. This leads to faults accumulating to a certain level before suddenly erupting, causing severe disruptions to network services. Summary of the Invention
[0004] This invention provides a network equipment fault prediction method, system, and medium based on artificial intelligence to solve the technical problems of high uncertainty caused by signal discreteness in existing network equipment fault prediction methods, which leads to low equipment fault prediction accuracy and lagging maintenance management. It achieves the technical effect of improving fault prediction accuracy and operation and maintenance efficiency through entropy reduction decision-making and intelligent joint decision-making.
[0005] In a first aspect, the present invention provides a network device fault prediction method based on artificial intelligence, wherein the network device fault prediction method based on artificial intelligence includes:
[0006] By collecting network device operating signals, and based on the first judgment component embedded in the network peripheral interface, a fault entropy determination based on discrete signal points is performed to generate a fault prediction instruction and encapsulate network device data. According to the fault prediction instruction, the integrated central control feedback of the network device data is triggered, driving the predictive agent to perform joint decision-making based on physical laws and causal logic to determine the fault prediction result. Based on the entropy reduction decision based on fault entropy, a fault control scheme is determined. Based on the fault prediction result and the fault control scheme, fault operation and maintenance management is performed on the network device.
[0007] Secondly, the present invention also provides an artificial intelligence-based network device fault prediction system, wherein the artificial intelligence-based network device fault prediction system includes:
[0008] Fault Entropy Determination Module: By collecting network device operating signals, and based on the first judgment component embedded in the network peripheral interface, performs fault entropy determination based on discrete signal points, generates fault prediction instructions, and encapsulates network device data; Joint Decision Module: Based on the fault prediction instructions, triggers the integrated central control feedback of the network device data, drives the predictive agent to perform joint decision-making based on physical laws and causal logic, determines the fault prediction result, and determines the fault control scheme based on entropy reduction decision; Fault Operation and Maintenance Module: Performs fault operation and maintenance management on the network device based on the fault prediction result and the fault control scheme.
[0009] Thirdly, the present invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the network device fault prediction method based on artificial intelligence provided by the present invention.
[0010] This invention discloses a network device fault prediction method, system, and medium based on artificial intelligence, comprising: collecting network device operating signals; performing fault entropy determination based on discrete signal points according to a first judgment component embedded in the network peripheral interface; generating a fault prediction instruction and encapsulating network device data; triggering integrated central control feedback of the network device data according to the fault prediction instruction; driving a predictive agent to perform joint decision-making based on physical laws and causal logic to determine the fault prediction result; determining a fault control scheme based on entropy reduction decision; and performing fault operation and maintenance management on the network device according to the fault prediction result and the fault control scheme. The network device fault prediction method, system, and medium disclosed in this invention solve the technical problem of high uncertainty caused by signal discreteness in existing network device fault prediction methods, resulting in low equipment fault prediction accuracy and lagging maintenance management, and achieve the technical effect of improving fault prediction accuracy and operation and maintenance efficiency through entropy reduction decision and intelligent joint decision. Attached Figure Description
[0011] Figure 1 This is a flowchart illustrating the network device fault prediction method based on artificial intelligence according to the present invention.
[0012] Figure 2 This is a schematic diagram of the network device fault prediction system based on artificial intelligence according to the present invention.
[0013] Figure labeling: Fault entropy determination module 11, joint decision-making module 12, fault operation and maintenance module 13. Detailed Implementation
[0014] The above technical solutions will now be described in detail with reference to the accompanying drawings and specific embodiments to provide a better understanding of them. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. It should be understood that the present invention is not limited to the exemplary embodiments used only to explain the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention. Furthermore, it should be noted that, for ease of description, only the parts related to the present invention are shown in the drawings, not all of them.
[0015] Example 1, as Figure 1 This is a flowchart illustrating the network device fault prediction method based on artificial intelligence of the present invention, wherein the network device fault prediction method based on artificial intelligence includes:
[0016] By collecting network device operating signals, and based on the first judgment component embedded in the network peripheral interface, fault entropy judgment based on discrete signal points is performed, fault prediction instructions are generated, and network device data is encapsulated.
[0017] Specifically, the system first collects real-time operational signals from network devices to obtain their operating status and performance data. These signals include, but are not limited to, various modal signals such as voltage, current, temperature, network traffic, and bandwidth usage, reflecting the device's health status and providing foundational data for subsequent fault prediction. Next, the collected operational signals are frequency-localized at preset intervals to identify discrete signals, each representing the device's state at a specific moment. Then, a first judgment component pre-embedded in the network peripheral interface performs fault entropy determination on these discrete signals. If the calculated entropy value is greater than or equal to a preset fault entropy, it indicates a potential fault or anomaly in the network device. In this case, the first judgment component generates a fault prediction instruction to drive the predictive agent to execute subsequent joint decisions. Finally, the data generated from the fault entropy determination is encapsulated to form network device data for subsequent fault prediction and early warning, improving the accuracy and timeliness of fault prediction.
[0018] In some embodiments, the construction of the first determination component, based on a first determination component embedded in the network peripheral interface, includes:
[0019] For the target network device, read the cross-modal early degradation signal set; scan the cross-modal early degradation signal set, mine the fault entropy value under multi-level degradation, and determine the fault entropy sequence, wherein each sequence node represents the multi-signal mode comprehensive entropy value under different degradation levels; mine the entropy distribution pattern between single modes and define the fault mode guide; based on the fault entropy sequence and the fault mode guide, construct the first judgment component, open an external interface on the transmission network device side of the network device, and embed the first judgment component.
[0020] Specifically, the process begins by reading early degradation signals from the target network device under different operating conditions. These signals typically include multi-modal signals such as device voltage, temperature, network traffic, CPU utilization, memory usage, and packet transmission rate. These signals can be acquired in real time using embedded sensors or other monitoring tools, forming a complete cross-modal early degradation signal set. This set reflects the changes in multiple dimensions of the device as it transitions from normal operation to early degradation. Subsequently, the cross-modal early degradation signal set is scanned and analyzed. The set is then time-sequentially divided according to the degree of degradation, forming multiple segments of cross-modal early degradation signals. Each segment represents a degradation period, and these degradation periods are categorized into multiple degradation levels from low to high severity, such as normal, mild degradation, severe degradation, and fault state. For each segment of cross-modal early degradation signal, it is divided into modal signal sets. For each modal signal set, the signal value range is divided into several small intervals, each interval being considered a possible state. The probability of the signal value falling within each interval is then calculated, and this probability is used in the formula for calculating information entropy, such as Shannon entropy or Renyi entropy, to calculate the fault entropy value of each modal signal. By weighting the fault entropy values corresponding to each modal signal of each cross-modal early degradation signal, the multi-signal modal comprehensive entropy value of each cross-modal early degradation signal can be obtained. By mapping these multi-signal modal comprehensive entropy values to the degradation level corresponding to each cross-modal early degradation signal, a fault entropy sequence is formed as a sequence node. This fault entropy sequence can clearly characterize each stage of the equipment degradation process and the signal change trend corresponding to different degradation stages, providing basic data for subsequent fault diagnosis and prediction. Next, based on the fault entropy value of each single mode, the entropy distribution pattern of each single mode is mined to define a fault mode guide, that is, the correlation pattern and entropy distribution law between different signal modes when a device malfunctions. Then, based on the fault entropy sequence and fault mode guide, a first judgment component is constructed. This first judgment component can monitor the operating status of the device in real time based on the fault entropy sequence and fault mode guide, determine whether the device is in a normal state or has entered a degraded state, and promptly detect potential fault risks. Finally, in order to perform efficient real-time judgment and fault prediction, an external interface is opened on the transmission network device side of the device. Through this external interface, the first judgment component is deployed in an embedded manner in the device's hardware or software system. In this way, the operating signals of the network device can be transmitted to the first judgment component in real time for analysis and judgment, accurately predicting network device faults and providing a basis for decision-making for subsequent operation and maintenance management and fault repair.
[0021] In some embodiments, the entropy distribution pattern is defined by at least structural entropy, behavioral entropy, and environmental entropy.
[0022] Specifically, entropy distribution patterns can be categorized into at least structural entropy, behavioral entropy, and environmental entropy. Structural entropy primarily originates from factors such as the device's network structure, load, and configuration complexity. It measures system stability by analyzing the network topology, inter-device dependencies, network topology complexity, and device configuration complexity. For example, if a device has highly dependent nodes in the network or its configuration is complex, it may be more susceptible to certain failure factors, leading to system instability. Higher structural entropy indicates a more unstable or complex operational structure and a higher potential failure risk. The structural entropy is typically obtained by weighting the fault entropy values of structure-related single-mode faults. Behavioral entropy reflects the volatility of performance indicators and the density of error messages in the operational logs during device operation. For example, fluctuations in CPU utilization, memory usage, and network traffic are all part of behavioral entropy. Frequent fluctuations in these indicators or frequent log errors indicate unstable behavioral characteristics. This volatility increases the uncertainty of the device's state, affecting its overall operational reliability. The behavioral entropy is also typically obtained by weighting the fault entropy values of behavior-related single-mode faults. Environmental entropy reflects the stability of equipment in its operating environment and is typically measured using environmental data collected by sensors. Factors such as temperature fluctuations, voltage stability, and humidity changes in the environment can all affect the normal operation of the equipment. Fluctuations in environmental conditions like temperature and voltage directly impact hardware performance, potentially leading to malfunctions. The environmental entropy of a device is usually obtained by weighting the fault entropy values of single-mode faults related to the environment. After obtaining the device's structural entropy, behavioral entropy, and environmental entropy, the relationship between these values and corresponding thresholds can be used to determine fault mode orientations. For example, when both structural and behavioral entropy show high values exceeding their respective thresholds, it indicates potential fault risks in both hardware structure and performance. If the environmental entropy is high, and the behavioral entropy also exhibits abnormal fluctuations, it can be inferred that the device may be affected by external environmental factors, leading to performance instability and increasing the likelihood of failure. In summary, by comprehensively analyzing structural entropy, behavioral entropy, and environmental entropy, we can accurately assess the overall status of network devices, promptly identify potential problems, and reduce the risk of failure through subsequent proactive intervention and operation and maintenance management, ensuring that devices operate reliably in a stable, low-entropy state for a long time.
[0023] In some embodiments, performing fault entropy determination based on discrete signal points includes:
[0024] The system receives the network device operation signal, which includes a multi-mode signal; it traverses the network device operation signal according to a preset periodic interval to locate the signal frequency and determine a discrete signal sequence, wherein the discrete signal sequence corresponds one-to-one with the device operation signal; and it performs a priori determination on the discrete signal sequence according to the first determination component.
[0025] Specifically, the system first receives real-time operational signals from the target network device. These signals contain multiple modal data, reflecting the device's current health status. Then, the system iterates through and processes the network device's operational signals according to preset periodic intervals. During this process, for each modal signal, frequency statistics are performed at each preset periodic interval to identify the highest frequency value, which is then used as a discrete point within that interval. These discrete points are sorted chronologically to form a discrete signal sequence. Each discrete signal sequence corresponds one-to-one with a modal operational signal of the network device, representing the device's health status within that time window. Next, a first judgment component performs a priori judgment on each discrete signal sequence, calculating the corresponding fault entropy value and the total comprehensive entropy value to determine the current operational status of the network device. This allows for early identification of anomalies in device operation, reducing the need for manual intervention and improving the accuracy and timeliness of fault prediction.
[0026] In some embodiments, prior determination of the discrete signal sequence includes:
[0027] For the discrete signal sequence, fault entropy values are calculated sequence by sequence to determine multimodal entropy values and measure the overall signal entropy value. Based on the fault entropy sequence, the overall entropy value is matched. If the overall entropy value is less than the fault entropy of the minimum degradation level, the prediction process is terminated and a normal equipment identifier is generated.
[0028] Specifically, after obtaining the discrete signal sequence for each modal signal, these discrete signal sequences are input into the first judgment component. The first judgment component calculates the fault entropy value for each discrete signal sequence using the same method described above for calculating the fault entropy value of each modal signal in the cross-modal early degradation signal set, obtaining the fault entropy value for the corresponding modal signal for each discrete signal sequence. These fault entropy values are then aggregated and stored to form a multimodal entropy value. This value is then weighted and summed to obtain a comprehensive entropy value, reflecting the overall complexity and degree of abnormality of the device at the current moment. This comprehensive entropy value is then compared with the fault entropy sequence in the first judgment component to determine whether it is less than the fault entropy of the minimum degradation level, i.e., whether it falls within the normal range. If the comprehensive entropy value is within the normal level fault entropy range, the network device is determined to be in normal working condition, requiring no further warning or intervention measures. At this point, the prediction process is terminated, and a device normal status identifier is generated, indicating that the device is currently in a healthy state and requires no further operation, thereby improving the device's operational efficiency and stability.
[0029] In some embodiments, if the comprehensive entropy value is greater than or equal to the fault entropy of the minimum degradation level, the fault prediction instruction is generated, and the target fault entropy value based on the fault entropy sequence is located; for the multimodal entropy value, entropy distribution pattern matching is performed to determine the target fault direction; the signal feature vector of the discrete signal sequence, the target fault entropy value, and the target fault direction are encapsulated into network device data.
[0030] Specifically, if the calculated comprehensive entropy value is greater than or equal to the fault entropy of the minimum degradation level, it indicates that the network device has entered a fault warning state, and there may be a risk of failure or anomaly. At this time, a fault prediction instruction is generated. This instruction includes the current degradation level and is used to trigger subsequent data backhaul and joint decision-making. Subsequently, based on this comprehensive entropy value, the lowest-level node with the closest comprehensive entropy value is matched from the fault entropy sequence, and the fault entropy value of this node is defined as the target fault entropy value. This provides more accurate fault location for the fault prediction instruction, which is helpful for subsequent decision-making and adjustments. Afterwards, following the same analysis method described above, the entropy distribution pattern is matched based on the multimodal entropy value to calculate the corresponding structural entropy, behavioral entropy, environmental entropy, etc. Then, based on the relationship between this entropy value and the corresponding threshold, the target fault orientation is determined, providing guidance for subsequent fault control. Then, the key feature data (modes with higher fault entropy values) in each discrete signal sequence and the corresponding fault entropy values are concatenated according to a preset template vector to form a signal feature vector. This signal feature vector is then encapsulated with the target fault entropy value and the target fault direction to form a network device data packet. This network device data packet is sent to the central control platform for further fault diagnosis, early warning and control decisions to ensure the stability and reliability of the equipment.
[0031] According to the fault prediction instruction, the integrated control backhaul of the network device data is triggered, driving the predictive agent to perform joint decision-making based on physical laws and causal logic, determine the fault prediction result, and determine the fault control scheme based on the entropy reduction decision of fault entropy.
[0032] Specifically, after generating a fault prediction command, the network device's data feedback mechanism is triggered based on this command, transmitting the encapsulated network device data to the central control platform. The central control platform integrates and consolidates this data to support subsequent decision-making. Then, using the received signal feature vector as input data and the target fault orientation as a constraint, the central control platform drives the predictive agent to make joint decisions from both physical and causal logic dimensions, generating a fault prediction result. Afterward, using the target fault entropy value as a baseline, the fault prediction result is iteratively optimized to reduce entropy, formulating a corresponding fault control scheme for proactive fault intervention, reducing the frequency of device failures, and improving the reliability and operational efficiency of network devices.
[0033] In some embodiments, the construction of the predictive agent before driving it to perform joint decision-making based on physical laws and causal logic includes:
[0034] Based on the first physical law of network equipment operation and maintenance, a first network is mined, wherein the first physical law includes at least electron migration, heat conduction, and protocol state machine; based on the second causal logic of network equipment operation and maintenance, a second network is mined, wherein the second causal logic defines node elements at least as device components, software vulnerabilities, configuration policies, and network traffic behavior; the first network and the second network are merged to form a predictive intelligent agent.
[0035] Specifically, before making joint decisions, a predictive agent needs to be constructed, which includes a first network and a second network. The first network is established based on the first physical law of network equipment operation and maintenance. This first physical law is the basic physical characteristic of equipment operation, including at least electromigration, heat conduction, and protocol state machines. Electromigration describes how electrons move in the equipment during the flow of current. Different devices and components may experience overheating, electrical faults, etc., due to uneven electron flow under the influence of current. By using the state of each electronic component or circuit as nodes and the current flow path as edges, a graph describing electromigration can be constructed to infer and analyze the current and voltage characteristics of each part of the equipment and determine the possible electrical anomalies. Heat conduction refers to the process by which electrons move within the equipment. The propagation and distribution of internal heat, and the operating temperature of different devices, directly affect their performance. Excessive temperature may lead to device malfunction or reduced efficiency. By treating each component of the device as a node and the heat transfer path as an edge, a graph describing heat conduction can be constructed to identify potential thermal failure risks. A protocol state machine is a model of the state transitions of network devices under different network protocols. Devices may switch their operating states under different protocol modes (such as TCP, UDP, etc.). This protocol state machine describes the logic and behavior of such switching. By analyzing the state transitions of devices under different protocols, errors or faults in network protocols or data communication processes can be discovered. By paralleling and integrating the modeling results of various physical laws, a complete first network can be obtained, which can be used to analyze potential physical faults in devices.The second network is established based on the second causal logic of network equipment operation and maintenance. This second causal logic refers to the logical relationships and influencing factors of equipment failures, including at least equipment components, software vulnerabilities, configuration policies, and network traffic behavior. Equipment components are the hardware parts of the equipment, such as power modules, heat dissipation modules, CPUs, and memory. The performance and status of each component can affect the overall operation of the equipment. By using equipment components as nodes and the connections between components (such as data transmission and power consumption) as edges, a graph describing the equipment components can be constructed to identify fault propagation paths and potential risks between components. Software vulnerabilities are defects in the equipment's software system that can lead to system crashes, data leaks, and functional failures. Based on known software vulnerability databases and the actual operating conditions of the equipment, a graph related to software vulnerabilities can be constructed. A causal network graph is used to analyze the impact of software problems on device failures. Configuration strategies play a crucial role in the stability and performance of network devices. Improper configurations, such as unreasonable network bandwidth and incorrect routing policies, can lead to device failures or performance degradation. By treating each device configuration parameter as a node and the dependencies between configuration strategies as edges, a graph describing configuration strategies can be constructed to analyze the causal relationship between configuration errors and device failures. Network traffic behavior refers to abnormal network traffic behaviors, such as bandwidth overload, packet loss, and DDoS attacks, which can cause device failures. By treating each network traffic characteristic as a node and the relationship between abnormal network traffic and device failures as edges, a graph describing network traffic can be constructed to assess the impact of changes in traffic behavior on device performance and stability. By fusing and stitching these graphs together, a complete second network can be constructed to describe the causal relationships between various device components, configurations, traffic behaviors, and other factors, providing a basis for predicting device failures. After the construction of the first and second networks is completed, the two networks are connected in parallel and merged to form a comprehensive predictive agent. This predictive agent integrates physical laws and causal logic, and has a more comprehensive and accurate fault prediction capability, which can effectively avoid the occurrence of potential faults and improve the stability and reliability of the equipment.
[0036] In some embodiments, driving a predictive agent to perform joint decision-making based on physical laws and causal logic includes:
[0037] Based on the signal feature vector, node matching is performed in the first network, and directed reasoning is performed with the target fault guidance as a constraint to determine a first prediction result; node matching is performed in the second network, and directed reasoning is performed with the target fault guidance as a constraint to determine a second prediction result; the first prediction result and the second prediction result are concatenated as a fault prediction result; with the target fault entropy value as a baseline, optimization iterative decision-making based on entropy reduction guidance is performed on the fault prediction result to generate a fault control scheme, wherein the taboo list is locked with a preset entropy reduction scale and unlocked with a preset number of iterations.
[0038] Specifically, during joint decision-making, the signal feature vector is input into the predictive agent. The predictive agent uses key features from the signal feature vector to perform node matching on each physical node in the first network, obtaining first network nodes containing these key features. Similarly, it uses key features from the signal feature vector to perform node matching on the second network, obtaining second network nodes containing these key features. Subsequently, guided by the target fault orientation, the matched first and second network nodes perform directed inference to infer the possible fault type and entropy value of the device in its current state. For example, if the entropy value of the temperature signal increases significantly, the inference process may predict that the device is overheating; if the device's CPU utilization is correlated with abnormal fluctuations in network traffic, the inference process may point to the risk of misconfiguration or network traffic attack. After the inference is completed by the first and second network nodes, a first prediction result and a second prediction result can be obtained. By merging and concatenating the first and second prediction results, a comprehensive fault prediction result can be obtained. Next, using the target fault entropy value as a baseline, an optimization iterative decision based on entropy reduction is initiated. During this process, based on the fault prediction results, the comprehensive entropy value of the current equipment state is first calculated. This can be obtained by summing the entropy values of each node, representing the total uncertainty of the equipment. Then, based on the difference between this comprehensive entropy value and the target fault entropy value, an initial fault control scheme is set. This initial fault control scheme is the closest one matched from the sample schemes based on the difference, and the tabu list size and maximum number of iterations are set. Then, at the start of the first round of optimization, the comprehensive entropy value simulating the use of this initial fault control scheme is calculated. If the current comprehensive entropy value decreases by a preset amount compared to the previous comprehensive entropy value, the current solution is considered to have made sufficient improvement. In each iteration, a set of neighborhood solutions is generated based on the current solution. These neighborhood solutions are usually obtained by making small adjustments to the current solution to maintain a relative optimization direction, such as changing some operating parameters of the equipment or adjusting the equipment configuration. For each neighborhood solution, its corresponding comprehensive entropy value is calculated, and its entropy reduction compared to the current solution is evaluated. If the entropy reduction of a neighborhood solution exceeds a set entropy reduction scale, the solution is considered to have significant optimization potential and is added to the tabu list for locking. Locked solutions are excluded from future search processes to prevent repeated attempts on the same solution within a short period. To avoid getting trapped in local optima during the search process, solutions in the tabu list are periodically unlocked according to unlocking rules. That is, if a locked solution does not bring significant optimization within a preset number of iterations, the tabu list will remove the restriction on that solution according to predetermined rules, allowing it to be considered again in subsequent iterations. This ensures the diversity of the search process and avoids search limitations.When the entropy value is gradually reduced through multiple iterations and the baseline is reached, the final fault control scheme is generated based on the optimal neighborhood solutions and their entropy reduction magnitude, and output to the equipment for fault management and optimization, thereby improving the operating efficiency and stability of the equipment and reducing the risk of fault occurrence.
[0039] In summary, the network device fault prediction method based on artificial intelligence provided by this invention has the following technical effects:
[0040] By collecting network device operating signals, and based on the first judgment component embedded in the network peripheral interface, a fault entropy determination based on discrete signal points is performed to generate a fault prediction instruction and encapsulate network device data. According to the fault prediction instruction, the integrated central control feedback of the network device data is triggered, driving the predictive agent to perform joint decision-making based on physical laws and causal logic to determine the fault prediction result. Then, based on entropy reduction decision-making, a fault control scheme is determined. Based on the fault prediction result and the fault control scheme, fault operation and maintenance management is performed on the network device, thereby achieving the technical effect of improving fault prediction accuracy and operation and maintenance efficiency through entropy reduction decision-making and intelligent joint decision-making.
[0041] Example 2, as Figure 2 This is a schematic diagram of the network device fault prediction system based on artificial intelligence according to the present invention. For example, Figure 1 The flowchart of the network device fault prediction method based on artificial intelligence of the present invention can be illustrated as follows: Figure 2 The structure shown is implemented.
[0042] Based on the same concept as the AI-based network device fault prediction method in the embodiments described above, the present invention also provides an AI-based network device fault prediction system comprising:
[0043] Fault Entropy Judgment Module 11: By collecting network device operating signals, and based on the first judgment component embedded in the network peripheral interface, performs fault entropy judgment based on discrete signal points, generates fault prediction instructions, and encapsulates network device data; Joint Decision Module 12: Based on the fault prediction instructions, triggers the integrated central control feedback of the network device data, drives the predictive agent to perform joint decision-making based on physical laws and causal logic, determines the fault prediction result, and determines the fault control scheme based on the entropy reduction decision based on fault entropy; Fault Operation and Maintenance Module 13: Performs fault operation and maintenance management on the network device based on the fault prediction result and the fault control scheme.
[0044] In some embodiments, the fault entropy determination module 11 includes:
[0045] For the target network device, read the cross-modal early degradation signal set; scan the cross-modal early degradation signal set, mine the fault entropy value under multi-level degradation, and determine the fault entropy sequence, wherein each sequence node represents the multi-signal mode comprehensive entropy value under different degradation levels; mine the entropy distribution pattern between single modes and define the fault mode guide; based on the fault entropy sequence and the fault mode guide, construct the first judgment component, open an external interface on the transmission network device side of the network device, and embed the first judgment component.
[0046] In some embodiments, the fault entropy determination module 11 includes:
[0047] The entropy distribution pattern is defined by at least structural entropy, behavioral entropy, and environmental entropy.
[0048] In some embodiments, the fault entropy determination module 11 includes:
[0049] The system receives the network device operation signal, which includes a multi-mode signal; it traverses the network device operation signal according to a preset periodic interval to locate the signal frequency and determine a discrete signal sequence, wherein the discrete signal sequence corresponds one-to-one with the device operation signal; and it performs a priori determination on the discrete signal sequence according to the first determination component.
[0050] In some embodiments, the fault entropy determination module 11 includes:
[0051] For the discrete signal sequence, fault entropy values are calculated sequence by sequence to determine multimodal entropy values and measure the overall signal entropy value. Based on the fault entropy sequence, the overall entropy value is matched. If the overall entropy value is less than the fault entropy of the minimum degradation level, the prediction process is terminated and a normal equipment identifier is generated.
[0052] In some embodiments, the fault entropy determination module 11 includes:
[0053] If the comprehensive entropy value is greater than or equal to the fault entropy of the minimum degradation level, the fault prediction instruction is generated, and the target fault entropy value based on the fault entropy sequence is located; for the multimodal entropy value, entropy distribution pattern matching is performed to determine the target fault direction; the signal feature vector of the discrete signal sequence, the target fault entropy value, and the target fault direction are encapsulated into network device data.
[0054] In some embodiments, the joint decision-making module 12 includes:
[0055] Based on the first physical law of network equipment operation and maintenance, a first network is mined, wherein the first physical law includes at least electron migration, heat conduction, and protocol state machine; based on the second causal logic of network equipment operation and maintenance, a second network is mined, wherein the second causal logic defines node elements at least as device components, software vulnerabilities, configuration policies, and network traffic behavior; the first network and the second network are merged to form a predictive intelligent agent.
[0056] In some embodiments, the joint decision-making module 12 includes:
[0057] Based on the signal feature vector, node matching is performed in the first network, and directed reasoning is performed with the target fault guidance as a constraint to determine a first prediction result; node matching is performed in the second network, and directed reasoning is performed with the target fault guidance as a constraint to determine a second prediction result; the first prediction result and the second prediction result are concatenated as a fault prediction result; with the target fault entropy value as a baseline, optimization iterative decision-making based on entropy reduction guidance is performed on the fault prediction result to generate a fault control scheme, wherein the taboo list is locked with a preset entropy reduction scale and unlocked with a preset number of iterations.
[0058] In embodiment three, the present invention also provides a computer-readable storage medium that can be used to store software programs, computer-executable programs, and modules, such as the program instructions / modules corresponding to the artificial intelligence-based network device fault prediction method in the embodiments of the present invention, thereby realizing the above-mentioned artificial intelligence-based network device fault prediction method.
[0059] It should be understood that the embodiments disclosed in this invention and the above description enable those skilled in the art to implement this invention. However, this invention is not limited to the embodiments mentioned above. It should be understood that those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this invention, and should all be included within the protection scope of this invention.
Claims
1. A network device fault prediction method based on artificial intelligence, characterized in that, The method includes: By collecting network device operation signals, and based on the first judgment component embedded in the network peripheral interface, a fault entropy judgment based on discrete signal points is executed, a fault prediction instruction is generated, and network device data is encapsulated. According to the fault prediction instruction, the integrated control feedback of the network device data is triggered, driving the prediction agent to perform joint decision-making based on physical laws and causal logic, determine the fault prediction result, and determine the fault control scheme based on the entropy reduction decision of fault entropy. Based on the fault prediction results and fault control scheme, the network equipment is subjected to fault operation and maintenance management. According to the first judgment component embedded in the network peripheral interface, the construction of the first judgment component includes: For the target network device, read the cross-modal early degradation signal set; Scan the cross-modal early degradation signal set, mine the fault entropy value under multi-level degradation, and determine the fault entropy sequence, wherein each sequence node represents the multi-signal modal comprehensive entropy value under different degradation levels; Discover the entropy distribution patterns among single modes and define fault mode guidance; Based on the fault entropy sequence and the fault mode guide, the first judgment component is constructed, and an external interface is opened on the transmission network device side of the network device to embed the first judgment component. The entropy distribution pattern is defined by at least structural entropy, behavioral entropy, and environmental entropy; Perform fault entropy determination based on discrete signal points, including: Receive the network device operation signal, wherein the network device operation signal includes a multimodal signal; According to a preset periodic interval, the network device operation signals are traversed to locate the signal frequency and determine the discrete signal sequence, which corresponds one-to-one with the device operation signal. Based on the first judgment component, a priori judgment is made on the discrete signal sequence; Performing priori determination on the discrete signal sequence includes: For the discrete signal sequence, the fault entropy value is calculated sequence by sequence to determine the multimodal entropy value and measure the overall signal entropy value; Based on the fault entropy sequence, the comprehensive entropy value is matched. If the comprehensive entropy value is less than the fault entropy of the minimum degradation level, the prediction process is terminated and a normal device identifier is generated. If the comprehensive entropy value is greater than or equal to the fault entropy of the minimum degradation level, the fault prediction instruction is generated, and the target fault entropy value based on the fault entropy sequence is located. For the aforementioned multimodal entropy values, entropy distribution pattern matching is performed to determine the target fault guidance; The signal feature vector of the discrete signal sequence, the target fault entropy value, and the target fault guidance are encapsulated into network device data; Based on the comprehensive entropy value, the lowest-level node that is closest to the current comprehensive entropy value is matched from the fault entropy sequence, and the fault entropy value of that node is defined as the target fault entropy value.
2. The network device fault prediction method based on artificial intelligence as described in claim 1, characterized in that, Before driving the predictive agent to execute joint decisions based on physical laws and causal logic, the construction of the predictive agent includes: Based on the first physical law of network equipment operation and maintenance, the first network is explored, wherein the first physical law includes at least electron migration, heat conduction, and protocol state machine; Based on the second causal logic of network equipment operation and maintenance, the second network is explored, wherein the second causal logic defines node elements at least by device components, software vulnerabilities, configuration policies, and network traffic behavior. The first network and the second network are combined to form a predictive agent.
3. The network device fault prediction method based on artificial intelligence as described in claim 2, characterized in that, Driving predictive agents to perform joint decision-making based on physical laws and causal logic includes: Based on the signal feature vector, node matching is performed in the first network, and directed reasoning is performed with the target fault guidance as a constraint to determine the first prediction result; In the second network, node matching is performed, and directed reasoning is carried out with the target fault guidance as a constraint to determine the second prediction result; The first prediction result and the second prediction result are cascaded together to form the fault prediction result; Using the target fault entropy value as a baseline, an optimization iterative decision based on entropy reduction is performed on the fault prediction result to generate a fault control scheme, wherein the taboo list is locked with a preset entropy reduction scale and unlocked with a preset number of iterations.
4. A network device fault prediction system based on artificial intelligence, characterized in that, The system for implementing the AI-based network device fault prediction method according to any one of claims 1-3, the system comprising: Fault Entropy Determination Module: By collecting network device operating signals, and based on the first judgment component embedded in the network peripheral interface, it performs fault entropy determination based on discrete signal points, generates fault prediction instructions, and encapsulates network device data. Joint Decision Module: Based on the fault prediction instruction, it triggers the integrated control feedback of the network device data, drives the predictive agent to perform joint decision-making based on physical laws and causal logic, determines the fault prediction result, and determines the fault control scheme based on the entropy reduction decision of fault entropy. Fault Operation and Maintenance Module: Based on the fault prediction results and fault control scheme, perform fault operation and maintenance management on the network devices.
5. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the AI-based network device fault prediction method as described in any one of claims 1 to 3.