Network fault diagnosis system and method based on quantum computing and machine learning

Abstract

The invention discloses a network fault diagnosis system and method based on quantum computing and machine learning. The system comprises a data acquisition module, a preprocessing module, a quantum computing module, a machine learning module, an algorithm fusion module and a resource allocation module. The data acquisition module acquires state data of network nodes and links in real time; the preprocessing module preprocesses the collected data; the quantum calculation module quickly screens suspected fault data through a quantum screening algorithm, and performs frequency domain analysis and time sequence analysis; the machine learning module trains a fault classification model according to the historical data, and performs fault classification on the real-time data; the algorithm fusion module generates a final fault diagnosis result through a multi-level weighted fusion strategy in combination with the frequency domain features, the time sequence features, the fault types and the fault change trends; and the resource allocation module dynamically adjusts network resource allocation according to the diagnosis result. According to the invention, the efficiency and precision of fault diagnosis are significantly improved.

Description

Technical Field

[0001] The present invention relates to the field of communication technology, and in particular to a network fault diagnosis system and method based on quantum computing and machine learning. Background Art

[0002] With the rapid development of information technology, the scale of networks continues to expand, and network architectures are becoming more and more complex. Especially in large-scale distributed systems, the status of network nodes and links is crucial to the performance and stability of the entire system. However, traditional network fault detection and diagnosis methods often have problems such as poor real-time performance, low accuracy, and difficulty in fault location, making it difficult to cope with various challenges in complex network environments.

[0003] Traditional network fault detection methods usually rely on rule-based detection systems to analyze network status data through simple threshold judgments or static models. These methods can usually only respond after a network fault occurs, and are not accurate enough in identifying and locating the fault type. Therefore, in the face of a large number of complex network nodes and links, the efficiency and accuracy of these methods often cannot meet actual needs.

[0004] In recent years, with the rapid development of technologies such as artificial intelligence, quantum computing, and machine learning, data-driven intelligent fault detection and diagnosis methods have gradually become a research hotspot. Quantum computing has become a potential tool for optimizing network fault detection and diagnosis due to its powerful computing power when processing complex data. In addition, machine learning models, especially deep learning and recursive neural networks, can provide accurate fault prediction and classification by learning from historical data, thereby improving the real-time and accuracy of fault diagnosis. Although some methods based on the combination of quantum computing and machine learning have emerged in the prior art, they still face the following problems:

[0005] Data processing complexity: The amount of network failure data is large and complex. How to efficiently process and analyze network status data is an urgent problem to be solved.

[0006] Application limitations of quantum computing: Although quantum computing has great potential in theory, its practical application still faces technical challenges, especially in how to design quantum algorithms suitable for network fault diagnosis.

[0007] Fault diagnosis accuracy and real-time issues: Network faults usually occur suddenly and may manifest in a variety of different types. Traditional single models are often unable to accurately identify and diagnose a variety of fault types in a timely manner.

[0008] Challenges of algorithm fusion: Existing network fault detection systems often do not have an effective way to integrate multiple algorithms such as quantum computing and machine learning, resulting in poor diagnostic results for the overall system.

[0009] To sum up, how to combine quantum computing and machine learning technology to design an efficient, accurate, and real-time network fault diagnosis system based on quantum computing and machine learning has become a hot topic in current research. Summary of the invention

[0010] In view of the shortcomings of the prior art, the purpose of the present invention is to provide a network fault diagnosis system and method based on quantum computing and machine learning, which is used to integrate quantum computing and machine learning to perform network fault diagnosis, thereby significantly improving the efficiency and accuracy of fault diagnosis.

[0011] To achieve the above objectives, the present invention provides the following technical solutions: a network fault diagnosis system and method based on quantum computing and machine learning, comprising:

[0012] The data collection module is used to collect the network status data of each network node and link in real time, as well as to collect historical status data;

[0013] A preprocessing module, connected to the data acquisition module, for preprocessing each of the network status data and each of the historical status data to obtain preprocessed status data and preprocessed historical data;

[0014] A quantum computing module, connected to the preprocessing module, for quickly screening the preprocessing state data according to a preset quantum screening algorithm to obtain suspected fault data, performing frequency domain analysis on the suspected fault data according to quantum Fourier transform to obtain frequency domain features, and performing time feature analysis on the suspected fault data according to a quantum phase estimation algorithm to obtain time series features;

[0015] A machine learning module, connected to the preprocessing module, for respectively training a fault classification model according to each of the historical state data, and inputting the preprocessing state data into the fault classification model to obtain a corresponding fault type;

[0016] An algorithm fusion module is connected to the quantum computing module and the machine learning module respectively, and includes:

[0017] A primary fusion unit, used for performing preliminary weighted fusion on the frequency domain features, the time series features and the fault type to obtain a preliminary fusion result;

[0018] An intermediate fusion unit, connected to the primary fusion unit, is used to predict the fault change trend of the preliminary fusion result based on a preset recursive neural network, and to form an intermediate fusion result based on the frequency domain feature, the time series feature, the fault type and the fault change trend;

[0019] A high-level fusion unit, connected to the intermediate fusion unit, is used to dynamically adjust the frequency domain feature according to the fault type and the fault change trend, and generate a final diagnosis result according to the adjusted intermediate fusion result;

[0020] A resource allocation module is connected to the algorithm fusion module and is used to adjust the network resource allocation at each of the network nodes and the links according to the final diagnosis result.

[0021] Furthermore, it also includes an external detection module connected to the algorithm fusion module, and the external detection module includes:

[0022] A first detection unit, configured to detect multiple power parameters at each of the network nodes and the link in real time;

[0023] A second detection unit, used for detecting multiple device performance parameters at each of the network nodes and the links in real time;

[0024] a detection and calculation unit, connected to the first detection unit and the second detection unit, respectively, for inputting each of the power parameters into a preset power stability calculation formula to calculate a power network stability value, and inputting each of the equipment performance parameters into a preset initial aging calculation formula to calculate a network equipment aging score;

[0025] Then the algorithm fusion module also includes a fusion optimization unit connected to the advanced fusion unit, and the fusion optimization unit is used to obtain a weight optimization coefficient according to the power network stability value and the network equipment aging score;

[0026] The advanced fusion unit optimizes and adjusts the final diagnosis result according to the weight optimization coefficient.

[0027] Furthermore, the external detection module further comprises a third detection unit connected to the detection calculation unit, and the third detection unit is used to detect the equipment fault parameters at each of the network nodes and the link in real time;

[0028] The detection calculation unit updates the initial aging calculation formula according to each of the equipment fault parameters to obtain an optimized aging calculation formula, and the optimized aging calculation formula is used to calculate the network equipment aging score according to each of the equipment performance parameters and each of the equipment fault parameters.

[0029] Furthermore, the power parameters include voltage, current load, power frequency, battery health status and power harmonics, and the power stability calculation formula is configured as:

[0030]

[0031] Among them, S(t) is used to represent the power network stability value at time t, V(t) is used to represent the voltage at time t, I(t) is used to represent the current load at time t, is used to represent the power consumption at time t, f(t) is used to represent the power frequency at time t, H(t) is used to represent the battery health status at time t, and λ is used to represent the preset network stability attenuation coefficient. is used to represent the power supply harmonics at time t, α, β, γ are used to represent the preset first constant coefficient, the second constant coefficient and the third constant coefficient, δ is used to represent the preset stability attenuation constant, Ω k is used to represent the stability parameter of the kth network node, and N is used to represent the number of the network nodes.

[0032] Furthermore, the device failure parameters include failure rate and number of failures, and the device performance parameters include CPU usage, memory usage, hard disk health index, network throughput and network response time;

[0033] The optimized aging calculation formula is configured as:

[0034]

[0035] Wherein, L(t) is used to represent the aging score of the network device at time t, N i (t) is used to represent the number of failures at time t, P i (t) is used to represent the failure rate at time t, α i The fault weighting factor for the preset, β i Used to represent the preset failure frequency index, δ i is used to represent the preset failure rate adjustment index, U(t) is used to represent the CPU usage at time t, W(t) is used to represent the memory usage at time t, Ξ(t) is used to represent the network throughput at time t, Θ k (t) is used to represent the hard disk health index at time t, φ is used to represent the CPU and memory weighting factors, λ1 is used to represent the preset equipment aging decay rate, ξ is used to represent the preset stability decay factor, ρ is used to represent the preset dynamic weighting factor, σ′ is used to represent the network response time, and t0 is used to represent the preset start time.

[0036] Further, the external detection module includes a fourth detection unit, and the fourth detection unit is used to detect multiple external environmental parameters at each of the network nodes and the link in real time, and the external environmental parameters include ambient temperature, ambient humidity, ambient air pressure and ambient magnetic field;

[0037] The fusion optimization unit inputs the power network stability value, the network equipment aging score, the ambient temperature, the ambient humidity, the ambient air pressure and the ambient magnetic field into a preset weight optimization calculation formula to calculate a weight optimization coefficient.

[0038] Furthermore, the weight optimization calculation formula is configured as:

[0039]

[0040] Wherein, W(t) is used to represent the weight optimization coefficient, T i (t) is used to represent the ambient temperature at time t, H i (t) is used to represent the ambient humidity at time t, P i (t) is used to represent the ambient air pressure at time t, B i (t) is used to represent the environmental magnetic field at time t, α i ,β i , γ i , δ i , η i They are respectively used to represent the preset first temperature weighting factor, the second temperature weighting factor, the humidity weighting factor, the air pressure weighting factor and the magnetic field weighting factor, k is used to represent the preset exponential decay factor, ζ is used to represent the preset circuit stability adjustment coefficient, θ is used to represent the preset equipment aging adjustment factor, σ is used to represent the preset network throughput weighting factor, and t1 is used to represent the preset end time.

[0041] Furthermore, the resource allocation module includes:

[0042] A resource allocation unit, configured to identify the network topology to be repaired at each of the network nodes and the link according to the final diagnosis result, and readjust the network resource allocation at each of the network nodes and the link based on a preset quantum optimization algorithm;

[0043] A topology adjustment unit is connected to the resource allocation unit and is used to adjust the network topology structure according to the adjusted network resources.

[0044] Furthermore, the quantum optimization algorithm is a quantum annealing algorithm.

[0045] A network fault diagnosis method based on quantum computing and machine learning, applied to the above-mentioned network fault diagnosis system based on quantum computing and machine learning, comprising:

[0046] Step S1, the data collection module collects network status data at each network node and link in real time, and collects historical status data;

[0047] Step S2, the preprocessing module preprocesses each of the network status data and each of the historical status data to obtain preprocessed status data and preprocessed historical data;

[0048] Step S3, the quantum computing module quickly screens the preprocessed state data according to a preset quantum screening algorithm to obtain suspected fault data, performs frequency domain analysis on the suspected fault data according to quantum Fourier transform to obtain frequency domain features, and performs time feature analysis on the suspected fault data according to a quantum phase estimation algorithm to obtain time series features;

[0049] Step S4, the machine learning module is trained to obtain a fault classification model according to each of the historical state data, and the pre-processed state data is input into the fault classification model to obtain a corresponding fault type;

[0050] Step S5, the primary fusion unit performs preliminary weighted fusion on the frequency domain features, the time series features and the fault type to obtain a preliminary fusion result; the intermediate fusion unit predicts the fault change trend of the preliminary fusion result based on a preset recursive neural network, and forms an intermediate fusion result based on the frequency domain features, the time series features, the fault type and the fault change trend; the advanced fusion unit dynamically adjusts the frequency domain features according to the fault type and the fault change trend, and generates a final diagnosis result according to the adjusted intermediate fusion result;

[0051] Step S6: the resource allocation module adjusts the network resource allocation at each of the network nodes and the links according to the final diagnosis result.

[0052] Beneficial effects of the present invention:

[0053] The present invention realizes the rapid screening and accurate analysis of network status data by combining quantum computing module with machine learning module. The quantum screening algorithm can quickly identify suspected fault data in large-scale data, significantly improving the real-time performance and processing efficiency of fault detection. Combining quantum Fourier transform to perform frequency domain analysis on suspected fault data and quantum phase estimation algorithm to analyze time characteristics, it can comprehensively and accurately capture the fault characteristics in the network.

[0054] At the same time, the present invention effectively combines frequency domain features, time series features and fault types through a multi-level weighted fusion strategy of the algorithm fusion module. The primary fusion unit preliminarily fuses different features, the intermediate fusion unit predicts the fault change trend based on the recursive neural network, and the advanced fusion unit further dynamically adjusts the features and generates the final fault diagnosis results. Through this multi-level algorithm fusion, the present invention can achieve more accurate fault prediction and identification of fault evolution trends, thereby improving the robustness and accuracy of the overall diagnostic system.

[0055] In addition, through the integration of quantum computing and machine learning, the present invention can not only diagnose faults, but also predict the changing trend of network faults. By predicting the changing trend of faults through recursive neural networks, possible faults can be foreseen in advance, so that relevant personnel can take measures in advance. In addition, by setting up a resource allocation module, the present invention can dynamically adjust network resource allocation according to the final diagnosis results, effectively optimize network performance, and improve network stability and resource utilization. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 It is a schematic diagram of the structure of the network fault diagnosis system based on quantum computing and machine learning in the present invention;

[0057] Figure 2 It is a flowchart of the steps of the network fault diagnosis method based on quantum computing and machine learning in the present invention.

[0058] Figure numerals: 1. Data acquisition module; 2. Preprocessing module; 3. Quantum computing module; 4. Machine learning module; 5. Algorithm fusion module; 51. Primary fusion unit; 52. Intermediate fusion unit; 53. Advanced fusion unit; 54. Fusion optimization unit; 6. Resource allocation module; 61. Resource allocation unit; 62. Topology adjustment unit; 7. External detection module; 71. First detection unit; 72. Second detection unit; 73. Detection calculation unit; 74. Third detection unit; 75. Fourth detection unit. DETAILED DESCRIPTION

[0059] The present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. The same parts are represented by the same reference numerals. It should be noted that the words "front", "rear", "left", "right", "upper" and "lower" used in the following description refer to directions in the accompanying drawings, and the words "bottom surface" and "top surface", "inner" and "outer" refer to directions toward or away from the geometric center of a specific component, respectively.

[0060] Example 1, reference Figure 1, which is the first embodiment of the present invention, provides a network fault diagnosis system based on quantum computing and machine learning, which can realize the integration of quantum computing and machine learning for network fault diagnosis, and significantly improve the efficiency and accuracy of fault diagnosis, including:

[0061] Data collection module 1, used to collect network status data at each network node and link in real time, as well as collect historical status data;

[0062] The preprocessing module 2 is connected to the data acquisition module 1 and is used to preprocess each network state data and each historical state data to obtain preprocessed state data and preprocessed historical data;

[0063] The quantum computing module 3 is connected to the preprocessing module 2 and is used to quickly screen the preprocessing state data according to a preset quantum screening algorithm to obtain suspected fault data, and to perform frequency domain analysis on the suspected fault data according to quantum Fourier transform to obtain frequency domain features, and to perform time feature analysis on the suspected fault data according to a quantum phase estimation algorithm to obtain time series features;

[0064] The machine learning module 4 is connected to the preprocessing module 2 and is used to train the fault classification model according to each historical state data, and input the preprocessing state data into the fault classification model to obtain the corresponding fault type;

[0065] The algorithm fusion module 5 is connected to the quantum computing module 3 and the machine learning module 4 respectively, and includes:

[0066] The primary fusion unit 51 is used to perform preliminary weighted fusion of frequency domain features, time series features and fault types to obtain a preliminary fusion result;

[0067] The intermediate fusion unit 52 is connected to the primary fusion unit 51 and is used to predict the fault change trend of the preliminary fusion result based on a preset recursive neural network, and to form an intermediate fusion result based on frequency domain features, time series features, fault type and fault change trend;

[0068] The high-level fusion unit 53 is connected to the intermediate fusion unit 52 and is used to dynamically adjust the frequency domain characteristics according to the fault type and the fault change trend, and generate a final diagnosis result according to the adjusted intermediate fusion result;

[0069] The resource allocation module 6 is connected to the algorithm fusion module 5 and is used to adjust the network resource allocation at each network node and link according to the final diagnosis result.

[0070] Working principle of embodiment 1:

[0071] The data acquisition module 1 collects relevant network status data in real time by deploying sensors at each network node and link. All data is collected through distributed acquisition units and transmitted to the data center in a unified manner. The real-time network status data includes the node's bandwidth utilization, latency, packet loss rate, throughput, etc. The historical status data includes network status data at historical moments, fault records, maintenance records, etc.

[0072] The preprocessing module 2 is responsible for standardizing and cleaning the collected network status data and historical status data to remove noise data and abnormal data to ensure data quality. The data preprocessing process includes:

[0073] Data cleaning: remove irrelevant data, fill in missing data, and eliminate possible duplicate data.

[0074] Data standardization: Convert network status data from different sources into a unified format and normalize the data so that data from different network nodes can be compared on the same scale.

[0075] Anomaly detection: Use rules or statistical methods to identify and remove abnormal data points to ensure data accuracy and reliability.

[0076] The quantum computing module 3 performs rapid screening and analysis based on the pre-processed network status data through quantum screening algorithm, quantum Fourier transform and quantum phase estimation algorithm.

[0077] The machine learning module 4 uses the historical status data for training to obtain a fault classification model, and inputs the preprocessed network status data into the fault classification model to obtain the fault type.

[0078] The training process of the fault classification model includes: using historical network status data to train different machine learning models (such as support vector machines, decision trees, deep neural networks, etc.) for fault type classification. The machine learning model is trained with labeled historical data to continuously improve classification accuracy. The preprocessed network status data is input into the fault classification model, and the model outputs the type of fault based on this data, such as node failure, link failure, bandwidth bottleneck, etc. The identification of fault type is the basis for subsequent fusion analysis.

[0079] The algorithm fusion module 5 is the core part of the system. It combines the analysis results of the quantum computing module 3 and the machine learning module 4, integrates various information, and outputs the final fault diagnosis result. It includes three submodules: the primary fusion unit 51, the intermediate fusion unit 52 and the advanced fusion unit 53. The primary fusion unit 51 performs weighted fusion on the frequency domain features and time series features output by the quantum computing module 3 and the fault type output by the machine learning module 4 to obtain a preliminary fusion result. The fusion at this stage mainly relies on weighted average or other simple weighting strategies. The intermediate fusion unit 52 predicts the fault change trend of the preliminary fusion result based on the preset recursive neural network (RNN). The recursive neural network can capture the time dependence of the fault and predict the possible evolution trend of the fault. By combining the frequency domain features, time series features, fault type and its change trend, a more accurate intermediate fusion result is formed. The advanced fusion unit 53 dynamically adjusts the frequency domain features according to the fault type and fault change trend, and generates the final fault diagnosis result according to the adjusted intermediate fusion result. Adjust the intermediate fusion result. According to the fault type and change trend, the frequency domain characteristics of the output of the quantum computing module 3 are dynamically adjusted to enhance the accuracy of fault diagnosis.

[0080] Finally, the resource allocation module 6 reallocates resources to network nodes and links according to the final diagnosis result to ensure the stability and efficiency of the network.

[0081] This embodiment realizes the rapid screening and accurate analysis of network status data through the combination of quantum computing module 3 and machine learning module 4. The quantum screening algorithm can quickly identify suspected fault data in large-scale data, significantly improving the real-time performance and processing efficiency of fault detection. Combining the quantum Fourier transform to perform frequency domain analysis on suspected fault data and the quantum phase estimation algorithm to analyze the time characteristics, it is possible to comprehensively and accurately capture the fault characteristics in the network.

[0082] At the same time, this embodiment effectively combines frequency domain features, time series features and fault types through the multi-level weighted fusion strategy of the algorithm fusion module 5. The primary fusion unit 51 preliminarily fuses different features, the intermediate fusion unit 52 predicts the fault change trend based on the recursive neural network, and the advanced fusion unit 53 further dynamically adjusts the features and generates the final fault diagnosis result. Through this multi-level algorithm fusion, this embodiment can achieve more accurate fault prediction and identification of fault evolution trends, thereby improving the robustness and accuracy of the overall diagnosis system.

[0083] In addition, through the integration of quantum computing and machine learning, this embodiment can not only diagnose faults, but also predict the changing trend of network faults. By predicting the changing trend of faults through recursive neural networks, possible faults can be foreseen in advance, so that relevant personnel can take measures in advance. In addition, by setting up a resource allocation module 6, this embodiment can dynamically adjust network resource allocation according to the final diagnosis results, effectively optimize network performance, and improve network stability and resource utilization.

[0084] Embodiment 2 is the second embodiment of the present invention. Different from the previous embodiment, this embodiment provides an external detection module 7, which can further improve the accuracy of network fault diagnosis. It includes an external detection module 7 and a connection algorithm fusion module 5. The external detection module 7 includes:

[0085] A first detection unit 71, used for real-time detection of multiple power parameters at each network node and link;

[0086] A second detection unit 72, used for real-time detection of multiple device performance parameters at each network node and link;

[0087] The detection and calculation unit 73 is connected to the first detection unit 71 and the second detection unit 72, respectively, and is used to input each power parameter into a preset power stability calculation formula to calculate the power network stability value, and input each device performance parameter into a preset initial aging calculation formula to calculate the network device aging score;

[0088] The algorithm fusion module 5 further includes a fusion optimization unit 54 connected to the advanced fusion unit 53, and the fusion optimization unit 54 is used to obtain a weight optimization coefficient according to the power network stability value and the network equipment aging score;

[0089] The advanced fusion unit 53 optimizes and adjusts the final diagnosis result according to the weight optimization coefficient.

[0090] Working principle of embodiment 2:

[0091] The power parameters reflect the stability of the power system and the power supply quality of the network equipment, and can help the system detect the impact of power anomalies on network equipment failures. The equipment performance parameters reflect the aging of the network equipment. The detection and calculation unit 73 calculates the power network stability value (used to reflect the network stability of the network node) according to each power parameter, and calculates the network equipment aging score (used to reflect the equipment aging degree of the network node) according to each equipment performance parameter. The fusion optimization unit 54 obtains the weight optimization coefficient according to the power network stability value and the network equipment aging score, so that the advanced fusion unit 53 multiplies the weight optimization coefficient with the final diagnosis result to achieve the optimization adjustment of the final diagnosis result, and achieves the optimization adjustment of the final diagnosis result, avoiding the influence of the power network stability of the network node and the aging degree of the equipment, improving the calculation accuracy of the final diagnosis result, and further improving the accuracy of network fault diagnosis.

[0092] Preferably, the power parameters include voltage, current load, power frequency, battery health status and power harmonics, and the power stability calculation formula is configured as:

[0093]

[0094] Among them, S(t) is used to represent the power network stability value at time t, V(t) is used to represent the voltage at time t, I(t) is used to represent the current load at time t, is used to represent the power consumption at time t, f(t) is used to represent the power frequency at time t, H(t) is used to represent the battery health status at time t, and λ is used to represent the preset network stability attenuation coefficient. It is used to represent the power supply harmonics at time t, α, β, γ are used to represent the preset first constant coefficient, second constant coefficient and third constant coefficient, δ is used to represent the preset stability attenuation constant, Ω k It is used to represent the stability parameter of the kth network node, and N is used to represent the number of network nodes.

[0095] Specifically, in this embodiment, the value range of the power network stability value is [0, 1], where 0 indicates extreme instability and 1 indicates complete stability. The power network stability value is a dynamic quantity that changes with time and can be calculated in real time based on factors such as the system's voltage, current, power frequency, battery health status, and harmonics. Power harmonics represent harmonic components generated by nonlinear loads or imbalances in the power system. The network stability attenuation coefficient represents the decay rate of the power network stability over time. The first constant coefficient, the second constant coefficient, and the third constant coefficient are used to adjust the weights of the contribution of voltage, current load, power frequency, battery health status, and power harmonics to network stability. The stability attenuation constant is a constant that adjusts the attenuation of system stability and affects the weighted degree of the summation term. The stability parameter is used to measure the stability impact of the node. This formula integrates the impact of various important factors (voltage, current load, power frequency, battery health status, and power harmonics) in the power network on the stability of the power network, and integrates multiple time domain and space domain variables through integral and summation functions, so that the power network stability value can be accurately calculated.

[0096] Preferably, the external detection module 7 further includes a third detection unit 74 connected to the detection calculation unit 73, and the third detection unit 74 is used to detect device fault parameters at each network node and link in real time;

[0097] The detection calculation unit 73 updates the initial aging calculation formula according to each device fault parameter to obtain an optimized aging calculation formula, and the optimized aging calculation formula is used to calculate the aging score of the network device according to each device performance parameter and each device fault parameter.

[0098] Specifically, in this embodiment, the fault parameters generated after the network node device fault occurs will affect the network device aging score, thereby affecting the accuracy of fault diagnosis. The initial aging calculation formula is optimized and updated using the detected device fault parameters, the influence of the fault parameters on the device aging calculation is introduced, and the calculation accuracy of the network device aging score is improved.

[0099] Preferably, the device failure parameters include failure rate and failure times, and the device performance parameters include CPU usage, memory usage, hard disk health index, network throughput, and network response time;

[0100] The optimized aging calculation formula is configured as:

[0101]

[0102] Among them, L(t) is used to represent the aging score of the network equipment at time t, N i (t) is used to represent the number of failures at time t, P i (t) is used to represent the failure rate at time t, αi The fault weighting factor for the preset, β i Used to represent the preset failure frequency index, δ i It is used to represent the preset failure rate adjustment index, U(t) is used to represent the CPU usage at time t, W(t) is used to represent the memory usage at time t, Ξ(t) is used to represent the network throughput at time t, Θ k (t) is used to represent the hard disk health index at time t, φ is used to represent the CPU and memory weighting factors, λ1 is used to represent the preset device aging decay rate, ξ is used to represent the preset stability decay factor, ρ is used to represent the preset dynamic weighting factor, σ′ is used to represent the network response time, and t0 is used to represent the preset start time.

[0103] Specifically, in this embodiment, the value range of the device aging score is [0, 1], where 0 means that the device is completely healthy and not aged; 1 means that the device is completely aged and cannot be used. It reflects the impact of the number of failures and the failure rate on equipment aging. By weighting the failure weighting factor and the failure number index, it ensures that the impact of the number of failures changes nonlinearly and the failure rate is also weighted and adjusted. Indicates the effect of device load (CPU and memory usage) on aging, and decays it through the time term The accelerated aging of the device over time is simulated. Health Index section: Reflects the impact of the health status of all nodes on aging. The lower the health index of the node, the greater the aging score. Throughput part: The effect of network throughput changing over time on aging is described. The higher the throughput and the heavier the device load, the faster the aging rate.

[0104] Preferably, the external detection module 7 includes a fourth detection unit 75, which is used to detect multiple external environmental parameters at each network node and link in real time, and the external environmental parameters include ambient temperature, ambient humidity, ambient air pressure and ambient magnetic field;

[0105] The fusion optimization unit 54 inputs the power network stability value, the network equipment aging score, the ambient temperature, the ambient humidity, the ambient air pressure and the ambient magnetic field into a preset weight optimization calculation formula to calculate the weight optimization coefficient.

[0106] Specifically, in this embodiment, changes in the external environment will accelerate the aging of network node equipment, thereby affecting the fault diagnosis of the network node equipment. By introducing external environmental parameters to optimize and correct the weight optimization coefficients, the accuracy of network fault diagnosis can be further improved.

[0107] Preferably, the weight optimization calculation formula is configured as:

[0108]

[0109] Among them, W(t) is used to represent the weight optimization coefficient, T i (t) is used to represent the ambient temperature at time t, H i (t) is used to represent the ambient humidity at time t, P i (t) is used to represent the ambient air pressure at time t, B i (t) is used to represent the ambient magnetic field at time t, α i ,β i , γ i , δ i , η i They are respectively used to represent the preset first temperature weighting factor, the second temperature weighting factor, the humidity weighting factor, the air pressure weighting factor and the magnetic field weighting factor, κ is used to represent the preset exponential decay factor, ζ is used to represent the preset circuit stability adjustment coefficient, θ is used to represent the preset equipment aging adjustment factor, σ is used to represent the preset network throughput weighting factor, and t1 is used to represent the preset end time.

[0110] Specifically, in this embodiment, the weight optimization calculation formula integrates multiple factors (ambient temperature, humidity, air pressure, magnetic field, equipment aging, etc.), and through weighting, integration, exponential decay and other mechanisms, network fault optimization can be carried out dynamically and intelligently. It not only improves the calculation efficiency and operational flexibility, but also enhances the adaptability and robustness of the system, helping network managers to achieve better fault prevention and optimization decisions, thereby ensuring the stable operation and efficient maintenance of the network system.

[0111] Preferably, the resource allocation module 6 includes:

[0112] A resource allocation unit 61, configured to identify the network topology to be repaired at each network node and link according to the final diagnosis result, and readjust the network resource allocation at each network node and link based on a preset quantum optimization algorithm;

[0113] The topology adjustment unit 62 is connected to the resource allocation unit 61 and is used to adjust the network topology structure according to the adjusted network resources.

[0114] Preferably, the quantum optimization algorithm is a quantum annealing algorithm.

[0115] Specifically, in this embodiment, the resource allocation unit 61 implements network resource allocation and topology structure adjustment based on the quantum annealing algorithm, and optimizes the resource allocation of network nodes and links through the advantages of quantum computing, ensuring that the network can be quickly restored and maintain network performance after a network failure. The topology adjustment unit 62 receives the optimized resource allocation result from the resource allocation unit 61 and dynamically adjusts the network topology. The adjustment process includes optimizing routing paths, adjusting bandwidth allocation, and reallocating computing resources, etc., to ensure that network traffic is efficiently transmitted under the new resource allocation scheme.

[0116] A network fault diagnosis method based on quantum computing and machine learning is applied to the above-mentioned network fault diagnosis system based on quantum computing and machine learning, such as Figure 2 As shown, including:

[0117] Step S1, the data collection module 1 collects network status data at each network node and link in real time, and collects historical status data;

[0118] Step S2, the preprocessing module 2 preprocesses each network state data and each historical state data to obtain preprocessed state data and preprocessed historical data;

[0119] Step S3, the quantum computing module 3 quickly screens the preprocessed state data according to the preset quantum screening algorithm to obtain suspected fault data, performs frequency domain analysis on the suspected fault data according to the quantum Fourier transform to obtain frequency domain features, and performs time feature analysis on the suspected fault data according to the quantum phase estimation algorithm to obtain time series features;

[0120] Step S4, the machine learning module 4 trains the fault classification model according to each historical state data, and inputs the pre-processed state data into the fault classification model to obtain the corresponding fault type;

[0121] Step S5, the primary fusion unit 51 performs preliminary weighted fusion on the frequency domain features, time series features and fault types to obtain a preliminary fusion result; the intermediate fusion unit 52 predicts the fault change trend of the preliminary fusion result based on the preset recursive neural network, and forms an intermediate fusion result based on the frequency domain features, time series features, fault types and fault change trends; the advanced fusion unit 53 dynamically adjusts the frequency domain features according to the fault type and fault change trend, and generates a final diagnosis result according to the adjusted intermediate fusion results;

[0122] Step S6: the resource allocation module 6 adjusts the network resource allocation at each network node and link according to the final diagnosis result.

[0123] The above are only preferred embodiments of the present invention. The protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technicians in this technical field, some improvements and modifications without departing from the principle of the present invention should also be regarded as the protection scope of the present invention.

Claims

1. A network fault diagnosis system based on quantum computing and machine learning, characterized in that: include: A data collection module (1) is used to collect network status data of each network node and link in real time, as well as to collect historical status data; A preprocessing module (2), connected to the data acquisition module (1), for preprocessing each of the network status data and each of the historical status data to obtain preprocessed status data and preprocessed historical data; A quantum computing module (3) is connected to the preprocessing module (2) and is used to quickly screen the preprocessing state data according to a preset quantum screening algorithm to obtain suspected fault data, perform frequency domain analysis on the suspected fault data according to quantum Fourier transform to obtain frequency domain features, and perform time feature analysis on the suspected fault data according to a quantum phase estimation algorithm to obtain time series features; A machine learning module (4), connected to the preprocessing module (2), for training a fault classification model according to each of the historical state data, and inputting the preprocessing state data into the fault classification model to obtain a corresponding fault type; An algorithm fusion module (5) is connected to the quantum computing module (3) and the machine learning module (4) respectively, and comprises: A primary fusion unit (51) is used to perform preliminary weighted fusion on the frequency domain features, the time series features and the fault type to obtain a preliminary fusion result; An intermediate fusion unit (52), connected to the primary fusion unit (51), is used to predict the fault change trend of the preliminary fusion result based on a preset recursive neural network, and to form an intermediate fusion result based on the frequency domain feature, the time series feature, the fault type and the fault change trend; A high-level fusion unit (53), connected to the intermediate fusion unit (52), is used to dynamically adjust the frequency domain feature according to the fault type and the fault change trend, and generate a final diagnosis result according to the adjusted intermediate fusion result; A resource allocation module (6) is connected to the algorithm fusion module (5) and is used to adjust the network resource allocation at each of the network nodes and the links according to the final diagnosis result.

2. The network fault diagnosis system based on quantum computing and machine learning according to claim 1 is characterized in that: It also includes an external detection module (7) connected to the algorithm fusion module (5), and the external detection module (7) includes: A first detection unit (71) is used to detect multiple power parameters at each of the network nodes and the link in real time; A second detection unit (72), configured to detect in real time a plurality of device performance parameters at each of the network nodes and the links; A detection calculation unit (73), connected to the first detection unit (71) and the second detection unit (72), respectively, for inputting each of the power parameters into a preset power stability calculation formula to calculate a power network stability value, and for inputting each of the equipment performance parameters into a preset initial aging calculation formula to calculate a network equipment aging score; The algorithm fusion module (5) further comprises a fusion optimization unit (54) connected to the advanced fusion unit (53), wherein the fusion optimization unit (54) is used to obtain a weight optimization coefficient according to the power network stability value and the network equipment aging score; The advanced fusion unit (53) optimizes and adjusts the final diagnosis result according to the weight optimization coefficient.

3. The network fault diagnosis system based on quantum computing and machine learning according to claim 2 is characterized in that: The external detection module (7) further comprises a third detection unit (74), connected to the detection calculation unit (73), and the third detection unit (74) is used to detect device fault parameters at each of the network nodes and the link in real time; The detection calculation unit (73) updates the initial aging calculation formula according to each of the equipment fault parameters to obtain an optimized aging calculation formula, wherein the optimized aging calculation formula is used to calculate the network equipment aging score according to each of the equipment performance parameters and each of the equipment fault parameters.

4. The network fault diagnosis system based on quantum computing and machine learning according to claim 3 is characterized in that: The power parameters include voltage, current load, power frequency, battery health status and power harmonics, and the power stability calculation formula is configured as: Among them, S(t) is used to represent the power network stability value at time t, V(t) is used to represent the voltage at time t, I(t) is used to represent the current load at time t, is used to represent the power consumption at time t, f(t) is used to represent the power frequency at time t, H(t) is used to represent the battery health status at time t, and λ is used to represent the preset network stability attenuation coefficient. is used to represent the power supply harmonics at time t, α, β, γ are used to represent the preset first constant coefficient, the second constant coefficient and the third constant coefficient, δ is used to represent the preset stability attenuation constant, Ω k is used to represent the stability parameter of the kth network node, and N is used to represent the number of the network nodes.

5. The network fault diagnosis system based on quantum computing and machine learning according to claim 4 is characterized in that: The device failure parameters include failure rate and failure times, and the device performance parameters include CPU usage, memory usage, hard disk health index, network throughput and network response time; The optimized aging calculation formula is configured as: Wherein, L(t) is used to represent the aging score of the network device at time t, N i (t) is used to represent the number of failures at time t, P i (t) is used to represent the failure rate at time t, α i The fault weighting factor for the preset, β i Used to represent the preset failure frequency index, δ i is used to represent the preset failure rate adjustment index, U(t) is used to represent the CPU usage at time t, W(t) is used to represent the memory usage at time t, Ξ(t) is used to represent the network throughput at time t, Θ k (t) is used to represent the hard disk health index at time t, φ is used to represent the CPU and memory weighting factors, λ1 is used to represent the preset equipment aging decay rate, ξ is used to represent the preset stability decay factor, ρ is used to represent the preset dynamic weighting factor, σ′ is used to represent the network response time, and t0 is used to represent the preset start time.

6. The network fault diagnosis system based on quantum computing and machine learning according to claim 5 is characterized in that: The external detection module (7) comprises a fourth detection unit (75), the fourth detection unit (75) being used to detect multiple external environmental parameters at each of the network nodes and the link in real time, the external environmental parameters comprising ambient temperature, ambient humidity, ambient air pressure and ambient magnetic field; The fusion optimization unit (54) inputs the power network stability value, the network equipment aging score, the ambient temperature, the ambient humidity, the ambient air pressure and the ambient magnetic field into a preset weight optimization calculation formula to calculate a weight optimization coefficient.

7. The network fault diagnosis system based on quantum computing and machine learning according to claim 6 is characterized in that: The weight optimization calculation formula is configured as: Wherein, W(t) is used to represent the weight optimization coefficient, T i (t) is used to represent the ambient temperature at time t, H i (t) is used to represent the ambient humidity at time t, P i (t) is used to represent the ambient air pressure at time t, B i (t) is used to represent the environmental magnetic field at time t, α i , β i , γ i , δ i , η i They are respectively used to represent the preset first temperature weighting factor, the second temperature weighting factor, the humidity weighting factor, the air pressure weighting factor and the magnetic field weighting factor, κ is used to represent the preset exponential decay factor, ζ is used to represent the preset circuit stability adjustment coefficient, θ is used to represent the preset equipment aging adjustment factor, σ is used to represent the preset network throughput weighting factor, and t1 is used to represent the preset end time.

8. The network fault diagnosis system based on quantum computing and machine learning according to claim 1 is characterized in that: The resource allocation module (6) comprises: A resource allocation unit (61), configured to identify the network topology to be repaired at each of the network nodes and the link according to the final diagnosis result, and readjust the network resource allocation at each of the network nodes and the link based on a preset quantum optimization algorithm; A topology adjustment unit (62) is connected to the resource allocation unit (61) and is used to adjust the network topology structure according to the adjusted network resources.

9. The network fault diagnosis system based on quantum computing and machine learning according to claim 8 is characterized in that: The quantum optimization algorithm is a quantum annealing algorithm.

10. A network fault diagnosis method based on quantum computing and machine learning, applied to the network fault diagnosis system based on quantum computing and machine learning as claimed in any one of claims 1 to 9, characterized in that: include: Step S1, the data collection module (1) collects network status data at each network node and link in real time, and collects historical status data; Step S2, the preprocessing module (2) preprocesses each of the network state data and each of the historical state data to obtain preprocessed state data and preprocessed historical data; Step S3, the quantum computing module (3) quickly screens the preprocessed state data according to a preset quantum screening algorithm to obtain suspected fault data, performs frequency domain analysis on the suspected fault data according to quantum Fourier transform to obtain frequency domain features, and performs time feature analysis on the suspected fault data according to a quantum phase estimation algorithm to obtain time series features; Step S4, the machine learning module (4) is trained to obtain a fault classification model according to each of the historical state data, and the pre-processed state data is input into the fault classification model to obtain a corresponding fault type; Step S5, the primary fusion unit (51) performs preliminary weighted fusion on the frequency domain features, the time series features and the fault type to obtain a preliminary fusion result; the intermediate fusion unit (52) predicts the fault change trend of the preliminary fusion result based on a preset recursive neural network, and forms an intermediate fusion result based on the frequency domain features, the time series features, the fault type and the fault change trend; The advanced fusion unit (53) dynamically adjusts the frequency domain feature according to the fault type and the fault change trend, and generates a final diagnosis result according to the adjusted intermediate fusion result; Step S6, the resource allocation module (6) adjusts the network resource allocation at each of the network nodes and the links according to the final diagnosis result.

Images