An optical path fault prediction system and method based on machine learning

Abstract

The present invention discloses an optical path fault prediction system and method based on machine learning, specifically relating to the technical field of optical path fault prediction, and is used to solve the problem that it is difficult to effectively identify the precursor of optical fiber fusing during the existing high-power transmission process. It preliminarily screens the multi-channel real-time power data and temperature data by using the support vector machine algorithm to judge whether there are abnormal fluctuations. When detecting abnormal fluctuations that may trigger fusing, it evaluates the dynamic change characteristics of the non-uniform absorption mode and identifies the main characteristics of the multiple reflection interference mode. Joint feature extraction is performed on the features to generate multi-dimensional fusing risk parameters. A decision threshold is set based on the multi-dimensional fusing risk parameters, and the feature range overlapping with the normal power fluctuation is screened out through an adaptive rule, and the suspected abnormal hot spot area is marked. Finally, a fusing risk assessment result is generated, and an optical path warning signal is sent when the conditions are met.

Description

Technical Field

[0001] The present invention relates to the technical field of optical path fault prediction. More specifically, the present invention relates to an optical path fault prediction system and method based on machine learning. Background Art

[0002] Optical path fault prediction refers to analyzing the operating state and related parameters of optical signals in an optical communication network to identify potential problems that may cause optical path interruption or performance degradation in advance; in scenarios of high-power optical transmission by optical fibers, such as high-power optical communication systems (such as ultra-high-speed backbone networks), laser medicine, laser cutting, and other devices that require high-power optical transmission, as well as scenarios with extremely high requirements for the reliability of optical fibers; during these high-power transmission processes, optical fibers are extremely vulnerable to the influence of local defects or abnormal absorption, forming high-temperature regions, resulting in a decline in the physical properties of the optical fiber end face or even melting. "Optical fiber melting" is an extreme fault. When it occurs, the local high-temperature region advances forward, quickly melting the entire optical fiber, usually causing complete communication interruption instantly and being difficult to quickly repair or restore.

[0003] In the prior art, it is difficult to effectively identify the precursors of optical fiber melting during high-power optical transmission of optical fibers, which may lead to the instant melting of optical fibers and cause serious communication interruptions. Summary of the Invention

[0004] In order to overcome the above-mentioned defects of the prior art, embodiments of the present invention provide an optical path fault prediction system and method based on machine learning to solve the problems raised in the above background art.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] An optical path fault prediction method based on machine learning includes the following steps:

[0007] Using the support vector machine algorithm to preliminarily screen the multi-channel real-time power data and temperature data in the target optical fiber transmission system to determine whether there are abnormal fluctuations that may trigger melting;

[0008] When there are abnormal fluctuations that may trigger melting, analyze the spatial frequency distribution of the absorption hot spots on the optical fiber end face through two-dimensional wavelet transform to evaluate the dynamic change characteristics of the non-uniform absorption pattern; analyze the interference intensity distribution of multi-path scattering in the optical fiber cladding through the optical path tracing algorithm to identify the main characteristics of the multiple reflection interference pattern;

[0009] Joint feature extraction is performed on the dynamic change characteristics of the non-uniform absorption pattern and the main characteristics of the multiple reflection interference pattern to obtain multi-dimensional melting risk parameters;

[0010] Set the judgment threshold according to the multi-dimensional fuse risk parameters, screen out the characteristic range overlapping with the normal power fluctuation through the adaptive rule, and mark the suspected abnormal hot spot area;

[0011] Generate a risk assessment result that the target optical fiber transmission system is about to fuse for the suspected abnormal hot spot area, and send an optical path warning signal when the judgment threshold is met.

[0012] In a preferred embodiment, use the support vector machine algorithm to preliminarily screen the multi-channel real-time power data and temperature data in the target optical fiber transmission system to determine whether there are abnormal fluctuations that may trigger a fuse, including:

[0013] Synchronously collect multi-channel real-time power data and temperature data from the target optical fiber transmission system. The multi-channel real-time power data includes the optical power measurement values of each transmission channel in the optical fiber, and the temperature data includes the local temperature measurement values at the optical fiber transmission nodes;

[0014] Perform data preprocessing on the collected multi-channel real-time power data and temperature data to generate the cleaned multi-channel real-time power data and temperature data;

[0015] Construct a multi-dimensional feature space based on the cleaned multi-channel real-time power data and temperature data, and use the multi-channel real-time power data and temperature data as the input of the feature vector;

[0016] Use the pre-trained support vector machine model to classify the constructed multi-dimensional feature space to determine whether there are abnormal fluctuations. Among them, the support vector machine model is trained based on the historical optical fiber fuse precursor data, and the classification boundary is optimized by setting the abnormal sample label during the training process;

[0017] Output the result of the support vector machine model classification as the judgment information on whether there are abnormal fluctuations that may trigger a fuse in the target optical fiber transmission system.

[0018] In a preferred embodiment, analyze the spatial frequency distribution of the optical fiber end face absorption hot spot through two-dimensional wavelet transform to evaluate the dynamic change characteristics of the non-uniform absorption mode, including:

[0019] Collect real-time absorption hot spot distribution data on the target optical fiber end face. The absorption hot spot distribution data is recorded by a high-resolution infrared thermal imaging device at a specified sampling frequency. The recorded absorption hot spot distribution data includes the absorption hot spot intensity and spatial distribution characteristics at different positions on the optical fiber end face;

[0020] Perform two-dimensional wavelet transform processing on the absorption hot spot distribution data, convert the absorption hot spot distribution data from the time-space domain to the spatial frequency domain, and extract the high-frequency and low-frequency characteristics of different frequency bands;

[0021] Calculate the high-frequency energy concentration and low-frequency energy diffusivity of the absorption hotspots based on the two-dimensional wavelet transform results, where the high-frequency energy concentration characterizes the absorption non-uniformity of the hotspot region, and the low-frequency energy diffusivity characterizes the overall absorption distribution range of the hotspot region;

[0022] Package the high-frequency energy concentration and low-frequency energy diffusivity as the dynamic change characteristic data of the non-uniform absorption mode of the fiber end face.

[0023] In a preferred embodiment, the high-frequency energy concentration is calculated by the following formula: ; where, represents the high-frequency energy concentration, represents the total energy of the coefficients after two-dimensional wavelet transform in the high-frequency range, represents the threshold of the high-frequency component, represents the total energy in the entire frequency range;

[0024] The low-frequency energy diffusivity is calculated by the following formula: ; where, represents the low-frequency energy diffusivity, represents the total energy of the coefficients after two-dimensional wavelet transform in the low-frequency range.

[0025] In a preferred embodiment, when there is an abnormal fluctuation that may trigger a fuse, analyze the interference intensity distribution of multi-path scattering in the fiber cladding through the optical path tracing algorithm, and identify the main characteristics of the multi-reflection interference pattern, including:

[0026] Collect the intensity distribution data of the multi-path scattered light in the fiber cladding, and the intensity distribution data includes the interference intensity values of different paths and angles;

[0027] Preprocess the collected intensity distribution data in the fiber cladding, including removing environmental noise and abnormal signals and converting them into the path-angle matrix format;

[0028] Based on the path-angle matrix, use the optical path tracing algorithm to simulate the propagation process of the multi-path scattered light and generate the intensity distribution curve of the interference pattern;

[0029] Perform frequency-domain analysis on the intensity distribution curve of the interference pattern, and calculate the spectral characteristic index in the predetermined frequency range to quantify the main characteristics of the multi-reflection interference pattern.

[0030] In a preferred embodiment, the spectral characteristic index calculation formula is: ; where, represents the spectral characteristic index, defined as the energy integral in the predetermined frequency range, and respectively represent the upper and lower limits of the frequency range, Represents the energy density in the frequency domain.

[0031] In a preferred embodiment, joint feature extraction is performed on the dynamic change characteristics of the non-uniform absorption mode and the main characteristics of the multiple reflection interference mode to obtain multi-dimensional fuse risk parameters, including:

[0032] Obtain the high-frequency energy concentration, low-frequency energy diffusivity, and spectral characteristic index, and perform normalization processing to generate a multi-dimensional feature vector;

[0033] Perform joint feature extraction on the multi-dimensional feature vector, and use a convolutional neural network algorithm to extract correlation features;

[0034] Calculate the multi-dimensional fuse risk parameters for representing the optical fiber fuse risk based on the extracted correlation features.

[0035] In a preferred embodiment, set a decision threshold according to the multi-dimensional fuse risk parameters, and screen out the feature range overlapping with the normal power fluctuation through an adaptive rule to mark the suspected abnormal hot spot area, including:

[0036] Set an initial decision threshold based on the distribution characteristics of the multi-dimensional fuse risk parameters, and the initial decision threshold is calculated from the statistical distribution of the multi-dimensional fuse risk parameters;

[0037] Optimize the initial decision threshold according to the historical data of the optical fiber end face absorption hot spot and the dynamic change trend of the multi-dimensional fuse risk parameters to obtain the decision threshold;

[0038] Re-analyze the distribution characteristics of the optical fiber end face absorption hot spot, and calculate the local absorption intensity and its spatial distribution range;

[0039] Apply an adaptive rule to screen the re-analyzed absorption hot spot features, and eliminate the feature range overlapping with the normal power fluctuation to mark the suspected abnormal hot spot area.

[0040] In a preferred embodiment, generate a risk assessment result for the target optical fiber transmission system about to fuse for the suspected abnormal hot spot area, and send an optical path warning signal under the condition of meeting the decision threshold, including:

[0041] Obtain the multi-dimensional fuse risk parameters of the suspected abnormal hot spot area, and generate a fuse risk assessment result for the target optical fiber transmission system according to the comparison result between the multi-dimensional fuse risk parameters and the decision threshold;

[0042] Output the fuse risk assessment result and the optical path warning signal for subsequent optical path fault handling and execution of protection strategies.

[0043] On the other hand, the present invention provides an optical path fault prediction system based on machine learning, including a data screening module, a frequency analysis module, a scattering analysis module, a feature extraction module, a feature screening module, and a risk assessment module;

[0044] Data screening module: Using the support vector machine algorithm to preliminarily screen the multiplexed real-time power data and temperature data in the target optical fiber transmission system to determine whether there are abnormal fluctuations that may trigger fusing;

[0045] Frequency analysis module: When there are abnormal fluctuations that may trigger fusing, analyze the spatial frequency distribution of the absorption hot spots on the fiber end face through two-dimensional wavelet transform to evaluate the dynamic change characteristics of the non-uniform absorption mode;

[0046] Scattering analysis module: When there are abnormal fluctuations that may trigger fusing, analyze the interference intensity distribution of the multi-path scattering in the fiber cladding through the optical path tracing algorithm to identify the main characteristics of the multiple reflection interference mode;

[0047] Feature extraction module: Jointly extract features from the dynamic change characteristics of the non-uniform absorption mode and the main characteristics of the multiple reflection interference mode to obtain multi-dimensional fusing risk parameters;

[0048] Feature screening module: Set a judgment threshold according to the multi-dimensional fusing risk parameters, screen out the feature range overlapping with the normal power fluctuation through the adaptive rule, and mark the suspected abnormal hot spot area;

[0049] Risk assessment module: Generate a risk assessment result that the target optical fiber transmission system is about to fuse for the suspected abnormal hot spot area, and issue an optical path warning signal under the condition of meeting the judgment threshold.

[0050] Technical effects and advantages of the optical path fault prediction system and method based on machine learning of the present invention:

[0051] 1. By analyzing the multiplexed real-time power data and temperature data and combining the preliminary screening function of the support vector machine algorithm, it is possible to quickly determine whether there are abnormal fluctuations that may trigger fusing, providing a reliable basis for subsequent accurate analysis; when abnormal fluctuations are detected, the method further uses two-dimensional wavelet transform and optical path tracing algorithm to deeply analyze the absorption hot spot distribution on the fiber end face and the cladding scattering interference mode respectively, comprehensively capturing the dynamic characteristics of the non-uniform absorption mode and the multiple reflection interference mode; this joint analysis method not only improves the detection accuracy of the fusing precursor, but also significantly reduces the false positive rate, ensuring the accuracy and real-time performance of the optical path fault prediction.

[0052] 2. By jointly extracting the characteristics of the non-uniform absorption mode and the multiple reflection interference mode, multi-dimensional fuse risk parameters are generated, and the selected characteristic range is optimized by combining adaptive rules, effectively marking the suspected abnormal hot spots, realizing the accurate quantification of the fuse risk; finally, based on the evaluation results of the multi-dimensional fuse risk parameters, when the detected fuse risk exceeds the determination threshold, an optical path warning signal is sent in real time, providing important decision-making support for the operation and maintenance of the optical fiber communication network; the method of the present invention can not only significantly improve the accuracy and timeliness of optical path fault prediction, but also effectively reduce the communication interruption and equipment loss caused by optical fiber fusing, and improve the stability and reliability of the optical communication network. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 Schematic diagram of a method for predicting optical path faults based on machine learning according to the present invention;

[0054] Figure 2 Schematic diagram of the structure of a system for predicting optical path faults based on machine learning according to the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0055] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0056] Embodiment 1: Figure 1 A method for predicting optical path faults based on machine learning according to the present invention is given, which includes the following steps:

[0057] Use the support vector machine algorithm to preliminarily screen the multi-channel real-time power data and temperature data in the target optical fiber transmission system to determine whether there are abnormal fluctuations that may trigger fusing.

[0058] When there are abnormal fluctuations that may trigger fusing, analyze the spatial frequency distribution of the absorption hot spots at the optical fiber end face through two-dimensional wavelet transform to evaluate the dynamic change characteristics of the non-uniform absorption mode; analyze the interference intensity distribution of multi-path scattering in the optical fiber cladding through the optical path tracing algorithm to identify the main characteristics of the multiple reflection interference mode.

[0059] Jointly extract the dynamic change characteristics of the non-uniform absorption mode and the main characteristics of the multiple reflection interference mode to obtain multi-dimensional fuse risk parameters.

[0060] Set a determination threshold according to the multi-dimensional fuse risk parameters, and screen out the characteristic range overlapping with the normal power fluctuation through the adaptive rule to mark the suspected abnormal hot spot area.

[0061] Generate a risk assessment result for the target optical fiber transmission system about to fuse in the suspected abnormal hot spot area, and send an optical path warning signal when the judgment threshold is met.

[0062] Use the support vector machine algorithm to preliminarily screen the multi-channel real-time power data and temperature data in the target optical fiber transmission system to determine whether there are abnormal fluctuations that may trigger fusing, including:

[0063] Synchronously collect multi-channel real-time power data and temperature data from the target optical fiber transmission system. The multi-channel real-time power data includes the optical power measurement values of each transmission channel in the optical fiber, and the temperature data includes the local temperature measurement values at the optical fiber transmission nodes:

[0064] Among them, the optical power measurement value of each transmission channel in the optical fiber transmission system is recorded by the distributed optical power monitoring device at a millisecond-level sampling frequency; the local temperature measurement value of the surrounding environment of the optical fiber transmission node is collected in real time by the distributed temperature sensing device and synchronously recorded with the sampling time of the optical power data; during the data collection process, the sampling time points are ensured to be consistent through the clock synchronization mechanism.

[0065] Perform data preprocessing on the collected multi-channel real-time power data and temperature data to generate the cleaned multi-channel real-time power data and temperature data:

[0066] The data preprocessing of the collected multi-channel real-time power data and temperature data includes:

[0067] Noise removal: Use the median filtering algorithm to remove the high-frequency noise that may be introduced during the measurement process, such as abnormal signals caused by instantaneous electromagnetic interference.

[0068] Missing value filling: Fill in the missing data caused by equipment failures or signal interruptions in the data collection through the nearest neighbor interpolation algorithm to ensure the integrity of the data.

[0069] Normalization processing: Perform normalization processing on the multi-channel real-time power data and temperature data to standardize the data within a fixed range to reduce the impact of dimensional differences on subsequent analysis.

[0070] After the preprocessing is completed, generate the cleaned multi-channel real-time power data and temperature data.

[0071] Construct a multi-dimensional feature space based on the cleaned multi-channel real-time power data and temperature data, and use the multi-channel real-time power data and temperature data as the input of the feature vector:

[0072] Feature selection: Use the optical power measurement value of each optical fiber transmission channel and its corresponding local temperature measurement value as features to construct a feature vector containing multiple dimensions.

[0073] Feature combination: Through the combination of feature vectors, a multi-dimensional feature space jointly defined by multi-channel real-time power data and temperature data is formed, where each feature point represents the system operating state at a certain time point.

[0074] The constructed multi-dimensional feature space is used as input to a support vector machine model for classification.

[0075] Use a pre-trained support vector machine model to classify the constructed multi-dimensional feature space to determine whether there are abnormal fluctuations. Among them, the support vector machine model is trained based on historical precursor data of optical fiber fusing. During the training process, the classification boundary is optimized by setting abnormal sample labels:

[0076] Among them, the used support vector machine model is trained based on historical precursor data of optical fiber fusing. The training data includes multi-channel real-time power data and temperature data of a large number of optical fiber transmission systems in normal and abnormal states.

[0077] During the training process, by setting specific labels for abnormal samples, the classification boundary of the support vector machine model is adjusted so that the model can accurately distinguish between normal states and abnormal fluctuations that may trigger fusing.

[0078] The support vector machine model classifies each feature point in the multi-dimensional feature space and outputs the classification result.

[0079] Output the result of the support vector machine model classification as the determination information on whether there are abnormal fluctuations in the target optical fiber transmission system that may trigger fusing:

[0080] According to the classification result of the support vector machine model, generate the determination information on whether there are abnormal fluctuations in the target optical fiber transmission system that may trigger fusing. The determination information is represented in the form of a classification label. If it is determined that there are abnormal fluctuations, the result is used as the input for further analysis in subsequent steps.

[0081] Here, the target optical fiber transmission system refers to an optical fiber communication system for transmitting high-power optical signals, including an optical fiber transmission path and related node devices, such as optical power monitoring devices, temperature sensors, optical amplifiers, etc.

[0082] When there are abnormal fluctuations that may trigger fusing, analyze the spatial frequency distribution of the absorption hot spots on the optical fiber end face through two-dimensional wavelet transform to evaluate the dynamic change characteristics of the non-uniform absorption pattern, including:

[0083] Collect real-time absorption hot spot distribution data on the target optical fiber end face. The absorption hot spot distribution data is recorded by a high-resolution infrared thermal imaging device at a specified sampling frequency. The recorded absorption hot spot distribution data includes the absorption hot spot intensity and spatial distribution characteristics at different positions on the optical fiber end face:

[0084] When collecting real-time absorption hot spot distribution data on the fiber end face, a high-resolution infrared thermal imaging device is used to continuously monitor the fiber end face.

[0085] The configuration of the infrared thermal imaging device can select a high-sensitivity infrared thermal imaging device with a micron-level spatial resolution to ensure the ability to capture tiny absorption hot spots on the fiber end face. The sampling frequency of the device is set to more than a thousand frames per second to ensure the accuracy of dynamic absorption hot spot data.

[0086] The collected absorption hot spot distribution data includes the hot spot intensity values at different positions on the fiber end face and the spatial distribution coordinates of the hot spots, forming a time series of the hot spot distribution; through high-precision clock synchronization technology, the collected hot spot intensity data and the corresponding timestamps are bound to ensure the consistency of the sampling time, and the original absorption hot spot distribution data containing hot spot intensity and spatial distribution information is generated.

[0087] Perform two-dimensional wavelet transform processing on the absorption hot spot distribution data, convert the absorption hot spot distribution data from the spatio-temporal domain to the spatial frequency domain, and extract the high-frequency and low-frequency features of different frequency bands:

[0088] Convert the collected absorption hot spot distribution data from a time series to a two-dimensional array, where the rows and columns correspond to the spatial coordinates of the fiber end face, and the values represent the hot spot intensity values at each coordinate point.

[0089] Use two-dimensional wavelet transform to convert the absorption hot spot distribution data from the spatio-temporal domain to the spatial frequency domain, and the transformation formula is: ; where, represents the coefficient after two-dimensional wavelet transform, reflecting the energy intensity of the signal in a specific frequency range (determined by and ); represents the spatial function of the absorption hot spot distribution, which is a two-dimensional signal that defines the hot spot intensity value at any position point on the fiber end face; represents the wavelet basis function, which defines the frequency characteristics at a specific scale and position; is the scaling parameter, representing the scale size of the frequency, corresponds to the high-frequency component, corresponds to the low-frequency component; is the translation parameter, representing the translation operation of the signal in space.

[0090] The selection of the wavelet basis function can be a two-dimensional Gaussian wavelet basis function, because it has a high sensitivity to spatial frequency changes and is suitable for capturing the local frequency characteristics of absorption hot spots.

[0091] Decompose the two-dimensional wavelet transform result into a high-frequency part and a low-frequency part, and extract the corresponding high-frequency and low-frequency features for subsequent calculations.

[0092] Calculate the high-frequency energy concentration and low-frequency energy diffusivity of the absorption hotspots based on the two-dimensional wavelet transform results, where the high-frequency energy concentration characterizes the absorption non-uniformity of the hotspot region, and the low-frequency energy diffusivity characterizes the overall absorption distribution range of the hotspot region:

[0093] The high-frequency energy concentration is calculated by the following formula: ; where represents the high-frequency energy concentration, defined as the energy proportion of the high-frequency components, and is used to quantify the strength of the absorption non-uniformity in the hotspot region; represents the total energy of the coefficients after the two-dimensional wavelet transform in the high-frequency range; represents the threshold of the high-frequency components, which is empirically set according to the characteristics of the fiber hotspot distribution and is used to distinguish the high-frequency and low-frequency parts; represents the total energy in the entire frequency range and is used for normalization.

[0094] The low-frequency energy diffusivity is calculated by the following formula: ; where represents the low-frequency energy diffusivity, defined as the energy proportion of the low-frequency components, and is used to quantify the overall absorption distribution range of the hotspot region; represents the total energy of the coefficients after the two-dimensional wavelet transform in the low-frequency range.

[0095] Package the high-frequency energy concentration and low-frequency energy diffusivity as the dynamic change characteristic data of the non-uniform absorption mode of the fiber end face:

[0096] Store the values of the high-frequency energy concentration and low-frequency energy diffusivity according to the time series to generate a time series file of the dynamic change characteristic data; add a timestamp and a sampling point identifier to each group of dynamic change characteristic data to ensure the spatio-temporal consistency of subsequent data analysis; store the packaged dynamic change characteristic data in a dedicated data structure to ensure compatibility with the input format of the subsequent feature extraction step.

[0097] The high-frequency energy concentration and low-frequency energy diffusivity evaluate the dynamic change characteristics of the non-uniform absorption mode by quantifying the local non-uniformity and the overall distribution range of the absorption hotspots respectively. A higher high-frequency energy concentration indicates a drastic change in the absorption intensity within the hotspot region and significant local non-uniformity, which may be an early signal of melting; while a higher low-frequency energy diffusivity indicates a wide distribution range of the hotspots and a tendency for the absorption hotspots to spread. Combining the dynamic changes of the two can comprehensively reflect the evolution process of the hotspots from concentration to diffusion, reveal the complexity of the absorption mode and its dynamic impact on the local characteristics of the fiber end face, and provide key features for subsequent risk prediction.

[0098] When there are abnormal fluctuations that may trigger a fuse, the interference intensity distribution of the fiber cladding multipath scattering is analyzed through the optical path tracing algorithm to identify the main features of the multiple reflection interference pattern, including:

[0099] Collect the intensity distribution data of multipath scattered light in the optical fiber cladding. The intensity distribution data includes the interference intensity values ​​of different paths and angles:

[0100] Distributed fiber optic sensors are deployed within the fiber cladding, and the sensors are able to simultaneously record the intensity values ​​of multipath scattered light in different spatial and angular ranges.

[0101] Interference intensity values ​​of different paths: the intensity of light when it propagates along different paths within the cladding.

[0102] Interference intensity values ​​at different angles: the interference intensity of light at different incident and scattering angles.

[0103] The sampling frequency can be set to thousands of frames per second, ensuring that the dynamic changes of multipath optical signals inside the optical fiber are captured.

[0104] Preprocess the acquired intensity distribution data in the optical fiber cladding, including removing environmental noise and abnormal signals and converting them into path-angle matrix format:

[0105] Removing environmental noise: Using a low-pass filter algorithm to filter out background noise, such as invalid signals caused by external electromagnetic interference. Removing abnormal signals: Using a threshold detection method, signal data with abnormally high or low intensity values ​​are removed. For example, the value of the optical signal intensity that exceeds the range of the device is considered abnormal.

[0106] The data is converted into a "path-angle matrix" format, where the rows of the matrix represent the path number of the optical signal, the columns represent the angle of the scattered light, and the value of the matrix element represents the interference intensity at that path and angle.

[0107] Based on the path-angle matrix, the light path tracing algorithm is used to simulate the propagation process of multi-path scattered light and generate the intensity distribution curve of the interference pattern:

[0108] A fiber cladding geometric model is established. The model adopts a three-dimensional cylindrical structure. The thickness, refractive index distribution of the fiber cladding and the incident parameters of the optical signal are determined by the physical properties of the optical fiber.

[0109] Simulate light propagation: The reflection and refraction behavior of light in the cladding is modeled, and its propagation distance and phase change on different paths are calculated. The light path tracing algorithm formula is: ;in, Represents light in three-dimensional space Strength on represents the reflection coefficient, which is the proportion of energy loss when light is reflected from the inner surface of the fiber cladding; represents the absorption coefficient of the optical signal, defined as the rate of energy attenuation caused by material absorption along the propagation path of light; represents the propagation path length of the light ray in the cladding; represents the phase change during light propagation, defined as the phase shift caused by path length changes or medium characteristics.

[0110] According to the propagation characteristics of each path, the calculated intensity values and phase values are combined to generate the intensity distribution curve of the interference pattern.

[0111] Perform frequency-domain analysis on the intensity distribution curve of the interference pattern, and calculate the spectral characteristic index within a predetermined frequency range to quantify the main characteristics of the multiple reflection interference pattern:

[0112] Convert the intensity distribution curve from the time domain or spatial domain to the frequency domain, and the spectral function is expressed as: ; where, represents the frequency-domain intensity distribution function, defined as the frequency distribution after Fourier transform; represents the intensity signal in the time domain, defined as the time function of the intensity distribution curve of the interference pattern; represents the frequency, defined as the independent variable in the Fourier transform; represents the time variable, which is the independent variable in the time domain; represents the imaginary unit, used to represent the phase information in the Fourier transform.

[0113] The calculation formula for the spectral characteristic index is: ; where, represents the spectral characteristic index, defined as the energy integral within a predetermined frequency range, used to quantify the spectral characteristics of the interference pattern; and represent the upper and lower limits of the frequency range respectively, determined by the typical frequency range of the scattered light in the fiber cladding, and used to define the integration range of the spectral characteristic index; represents the energy density in the frequency domain, defined as the square modulus value of.

[0114] The spectral characteristic index quantifies the energy distribution characteristics of multipath scattered light within a predetermined frequency range by integrating the frequency-domain energy of the interference pattern intensity distribution curve; the magnitude of the spectral characteristic index reflects the main characteristics of the multiple reflection interference pattern by quantifying the energy distribution characteristics of multipath scattered light within the fiber cladding in a specific frequency range. A larger spectral characteristic index indicates that the interference pattern has significant characteristics in the high-energy frequency band, which may correspond to complex multipath interference behavior; while a smaller spectral characteristic index indicates that the interference energy distribution is relatively uniform, the interference pattern is simple or the interaction between paths is weak, thereby quantifying the pattern complexity and energy distribution characteristics.

[0115] Joint feature extraction is performed on the dynamic change characteristics of the non-uniform absorption pattern and the main characteristics of the multiple reflection interference pattern to obtain multi-dimensional fuse risk parameters, including:

[0116] Obtain the high-frequency energy concentration, low-frequency energy diffusivity, and spectral characteristic index, and perform standardization processing to generate a multi-dimensional feature vector:

[0117] Standardization processing: To ensure the unity of parameter dimensions, normalization methods are used to process the high-frequency energy concentration, low-frequency energy diffusivity, and spectral characteristic index.

[0118] Combine the standardized high-frequency energy concentration, low-frequency energy diffusivity, and spectral characteristic index into a multi-dimensional feature vector in the form of: ; where represents the multi-dimensional feature vector.

[0119] Perform joint feature extraction on the multi-dimensional feature vector, and use the convolutional neural network algorithm to extract associated features:

[0120] To reveal the deep-level associations between the features, joint feature extraction is performed on the generated multi-dimensional feature vector using the convolutional neural network algorithm. The specific steps are as follows:

[0121] Input the multi-dimensional feature vector into the input layer of the convolutional neural network, where each dimension corresponds to an input node.

[0122] The convolutional neural network performs convolutional calculations on the input feature vector to extract the correlations between the features of each dimension. The feature extraction process includes a convolutional layer and an activation layer, and after the convolutional operation, it is processed by the non-linear activation function ReLU (Rectified Linear Unit).

[0123] The output of the hidden layer of the convolutional neural network is the associated feature matrix, representing the deep-level associated features between the high-frequency energy concentration, low-frequency energy diffusivity, and spectral characteristic index.

[0124] Calculate the multi-dimensional fusing risk parameters for representing the optical fiber fusing risk based on the extracted correlation features:

[0125] From the correlation feature matrix extracted by the convolutional neural network, calculate the multi-dimensional fusing risk parameters for quantifying the optical fiber fusing risk. According to the data of each dimension in the correlation feature matrix, calculate the multi-dimensional fusing risk parameters: ; where represents the multi-dimensional fusing risk parameter, which is used to characterize the overall risk of optical fiber fusing; represents the weight of the th dimension, which is obtained by model training and optimization; is set to be greater than 0; represents the eigenvalue of the th dimension in the correlation feature matrix; represents the index of the feature dimension; represents the total number of feature dimensions, that is, the total number of the high-frequency energy concentration degree, the low-frequency energy diffusion degree, and the spectral characteristic index (here it is 3).

[0126] The calculated multi-dimensional fusing risk parameters are used as the final output for subsequent evaluation and early warning of optical fiber fusing risks; the larger the value of the multi-dimensional fusing risk parameters, the higher the optical fiber fusing risk.

[0127] Set the determination threshold according to the multi-dimensional fusing risk parameters, and screen out the feature range overlapping with the normal power fluctuation through the adaptive rule, and mark the suspected abnormal hot spot areas, including:

[0128] Set the initial determination threshold based on the distribution characteristics of the multi-dimensional fusing risk parameters, and the initial determination threshold is obtained by calculating the statistical distribution of the multi-dimensional fusing risk parameters:

[0129] Extract a set of historical data samples from the multi-dimensional fusing risk parameters, and conduct statistical analysis on them, including calculating the mean value, standard deviation, and distribution interval.

[0130] According to the statistical analysis results, calculate the initial determination threshold through a formula, for example, set the threshold range based on the three-standard-deviation principle (that is: mean ± 3 times the standard deviation). The initial determination threshold provides the basis for preliminary screening and is used for optimization in subsequent steps.

[0131] Optimize the initial determination threshold according to the historical data of the optical fiber end face absorption hot spot and the dynamic change trend of the multi-dimensional fusing risk parameters to obtain the determination threshold:

[0132] The initial determination threshold usually cannot fully cover the risk characteristics in different scenarios, so it needs to be optimized based on the historical data of the optical fiber end face absorption hot spot and the dynamic change trend of the multi-dimensional fusing risk parameters.

[0133] Extract the distribution characteristics of absorption hotspots on the fiber end face under different risk levels, and combine with the characteristics of absorption hotspots that have been fused in actual cases. Analyze the time series data of multi-dimensional fuse risk parameters to identify the dynamic change rules, such as whether the risk parameters show an abnormal upward trend over time.

[0134] Based on historical data and dynamic trends, adjust the upper and lower limits of the initial judgment threshold to ensure that the judgment threshold can adapt to the characteristic changes in the complex operating environment.

[0135] Re-analyze the distribution characteristics of absorption hotspots on the fiber end face, and calculate the local absorption intensity and its spatial distribution range:

[0136] Calculation of local absorption intensity: Collect the absorption intensity values of different regions on the fiber end face, and calculate the average intensity, peak intensity and intensity gradient in the local area.

[0137] Evaluation of spatial distribution range: Divide the fiber end face into zones by a two-dimensional grid method for statistical analysis, and evaluate the distribution area and coverage of hotspots.

[0138] Capture of dynamic change characteristics: Record the change trends of hotspot intensity and distribution range over time, providing a reference for subsequent screening of characteristic ranges.

[0139] Apply adaptive rules to screen the re-analyzed absorption hotspot characteristics, and eliminate the characteristic ranges that overlap with normal power fluctuations to mark suspected abnormal hotspot areas:

[0140] After obtaining the re-analysis results, apply adaptive rules to screen the characteristic ranges of absorption hotspots to eliminate non-abnormal characteristics that may be caused by normal power fluctuations.

[0141] Define a set of adaptive screening rules based on the characteristics of absorption hotspots, including: Areas where the change amplitude of characteristic values is less than the set threshold are regarded as normal ranges; Areas where the distribution range does not exceed the historical normal distribution standard are eliminated.

[0142] Apply the adaptive rules to match each hotspot characteristic one by one, screen out the areas that overlap with normal power fluctuations, and mark the remaining areas after screening as suspected abnormal hotspot areas.

[0143] Generate a risk assessment result for the target optical fiber transmission system that is about to fuse for the suspected abnormal hotspot area, and issue an optical path warning signal under the condition of meeting the judgment threshold, including:

[0144] Obtain the multi-dimensional fuse risk parameters of the suspected abnormal hotspot area, and generate a fuse risk assessment result for the target optical fiber transmission system according to the comparison result between the multi-dimensional fuse risk parameters and the judgment threshold:

[0145] Obtain the multi-dimensional fusing risk parameters of the suspected abnormal hot spot area from the previous steps.

[0146] When the multi-dimensional fusing risk parameter is greater than the determination threshold, it indicates that the absorption non-uniformity of the hot spot area is significant, the hot spot distribution range is large, or the multi-path interference characteristics are complex, suggesting a high risk of optical fiber fusing in this area;

[0147] When the multi-dimensional fusing risk parameter is less than or equal to the determination threshold, it indicates that the characteristics of the hot spot area are within the normal range, with a low risk and no significant threat to the stability of the optical fiber transmission system.

[0148] Based on the comparison result, an example of the fusing risk assessment result is as follows:

[0149] Location of the hot spot area: End face coordinates of the optical fiber transmission system (x = 25μm, y = 40μm);

[0150] Multi-dimensional fusing risk parameter: 0.85 (greater than the determination threshold 0.7);

[0151] Risk level: High risk;

[0152] Description of dynamic characteristics: The distribution range of the absorption hot spot shows a diffusion trend, and the local non-uniformity increases;

[0153] Suggested measures: Generate an optical path warning signal, and it is recommended to reduce the power load of this optical path or perform optical fiber protection operations.

[0154] Output the fusing risk assessment result and the optical path warning signal for subsequent optical path fault handling and the execution of protection strategies:

[0155] The warning signal includes the identification information of the high-risk optical path, the dynamic change trend of the hot spot area, and the recommended handling scheme (such as adjusting the power load or implementing optical fiber protection). Send the risk assessment result and the warning signal to the optical path monitoring system and record them as historical data for subsequent optical path fault handling and the execution of protection strategies, and at the same time provide basic support for further optimizing the risk model and warning mechanism.

[0156] For example, the optical path warning signal can be prompted through the system alarm interface, and the content includes: "Optical path ID: 001 detects a high-risk hot spot, location coordinates (x = 25μm, y = 40μm), multi-dimensional fusing risk parameter = 0.85, please immediately check the optical fiber status and unload the power."

[0157] Example 2: This example introduces an optical path fault prediction system based on machine learning. Figure 2The structural schematic diagram of an optical path fault prediction system based on machine learning according to the present invention is given, including a data screening module, a frequency analysis module, a scattering analysis module, a feature extraction module, a feature screening module, and a risk assessment module.

[0158] Data screening module: Use the support vector machine algorithm to preliminarily screen the multiplexed real-time power data and temperature data in the target optical fiber transmission system to determine whether there are abnormal fluctuations that may trigger fusing.

[0159] Frequency analysis module: When there are abnormal fluctuations that may trigger fusing, analyze the spatial frequency distribution of the absorption hot spots on the optical fiber end face through two-dimensional wavelet transform to evaluate the dynamic change characteristics of the non-uniform absorption mode.

[0160] Scattering analysis module: When there are abnormal fluctuations that may trigger fusing, analyze the interference intensity distribution of the multi-path scattering in the optical fiber cladding through the optical path tracing algorithm to identify the main characteristics of the multi-reflection interference mode.

[0161] Feature extraction module: Jointly extract features from the dynamic change characteristics of the non-uniform absorption mode and the main characteristics of the multi-reflection interference mode to obtain multi-dimensional fusing risk parameters.

[0162] Feature screening module: Set a judgment threshold according to the multi-dimensional fusing risk parameters, and screen out the feature range overlapping with the normal power fluctuation through an adaptive rule to mark the suspected abnormal hot spot area.

[0163] Risk assessment module: Generate a risk assessment result that the target optical fiber transmission system is about to fuse for the suspected abnormal hot spot area, and issue an optical path warning signal when the judgment threshold is met.

[0164] The above formulas are all dimensionless and take their numerical calculations. The formulas are obtained by collecting a large amount of data for software simulation to get a formula closest to the actual situation. The preset parameters and threshold selection in the formulas are set by those skilled in the art according to the actual situation.

[0165] The above embodiments can be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the processes or functions described in the embodiments of the present application are generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center by wired (such as infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or a data center that contains one or more collections of available media. The available media can be magnetic media (such as floppy disks, hard disks, magnetic tapes), optical media (such as DVDs), or semiconductor media. The semiconductor media can be a solid-state drive.

[0166] Those of ordinary skill in the art will realize that the modules and algorithm steps of the examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or in a combination of computer software and electronic hardware. Whether these functions are executed in hardware or software depends on the specific application and design constraints of the technical solution. A professional technician can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of this application.

[0167] Those skilled in the art can clearly understand that the specific working processes of the systems, devices, and modules described above can refer to the corresponding processes in the foregoing method embodiments and will not be elaborated herein.

[0168] In the several embodiments provided in this application, it should be understood that the disclosed systems, devices, and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the modules is only a logical function division, and there can be other division methods in actual implementation. For example, multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed couplings or direct couplings or communication connections to each other can be through some interfaces, and the indirect couplings or communication connections of the devices or modules can be in electrical, mechanical, or other forms.

[0169] The module described as a separation component may or may not be physically separated. The component shown as a module may or may not be a physical module. It may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0170] In addition, in each embodiment of the present application, each functional module can be integrated into a processing module, or each module can exist physically alone, or two or more modules can be integrated into one module.

[0171] If the above functions are implemented in the form of software function modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in each embodiment of the present application. The aforementioned storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical discs that can store program codes.

[0172] The above is only the specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present application can easily think of changes or substitutions, which should all be covered by the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

[0173] Finally: The above is only the preferred embodiment of the present invention and is not used to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for predicting optical path faults based on machine learning, characterized in that: The steps include: Use the support vector machine algorithm to preliminarily screen the multi-channel real-time power data and temperature data in the target optical fiber transmission system to determine whether there are abnormal fluctuations that may trigger a fuse; When there are abnormal fluctuations that may trigger fusing, the spatial frequency distribution of the absorption hotspots at the fiber end face is analyzed by two-dimensional wavelet transform to evaluate the dynamic change characteristics of the non-uniform absorption mode, including: Collect real-time absorption hotspot distribution data on the target optical fiber end face. The absorption hotspot distribution data is recorded at a specified sampling frequency by a high-resolution infrared thermal imaging device. The recorded absorption hotspot distribution data includes the absorption hotspot intensity and spatial distribution characteristics at different positions on the optical fiber end face. Perform two-dimensional wavelet transform on the absorption hotspot distribution data, convert the absorption hotspot distribution data from the time-space domain to the spatial frequency domain, and extract high-frequency features and low-frequency features of different frequency bands; Based on the results of two-dimensional wavelet transform, the high-frequency energy concentration and low-frequency energy diffusion of the absorption hotspot are calculated, where the high-frequency energy concentration represents the absorption inhomogeneity of the hotspot area, and the low-frequency energy diffusion represents the overall absorption distribution range of the hotspot area; The high-frequency energy concentration and low-frequency energy diffusion are packaged as the dynamic change characteristic data of the non-uniform absorption mode of the optical fiber end face; When there are abnormal fluctuations that may trigger a fuse, the interference intensity distribution of the fiber cladding multipath scattering is analyzed through the optical path tracing algorithm to identify the main features of the multiple reflection interference pattern, including: Collecting the intensity distribution data of multi-path scattered light in the optical fiber cladding, the intensity distribution data includes interference intensity values ​​of different paths and angles; Preprocessing the collected intensity distribution data in the optical fiber cladding, including removing environmental noise and abnormal signals and converting them into a path-angle matrix format; Based on the path-angle matrix, the light path tracing algorithm is used to simulate the propagation process of multi-path scattered light and generate the intensity distribution curve of the interference pattern; Performing frequency domain analysis on the intensity distribution curve of the interference pattern, calculating the spectrum characteristic index within a predetermined frequency range to quantify the main characteristics of the multiple reflection interference pattern; The dynamic change characteristics of the non-uniform absorption pattern and the main features of the multiple reflection interference pattern are jointly extracted to obtain multi-dimensional fuse risk parameters; The judgment threshold is set according to the multi-dimensional fuse risk parameters, and the feature range overlapping with normal power fluctuations is screened out through adaptive rules to mark suspected abnormal hot spots; Generate risk assessment results of impending fuse failure in the target optical fiber transmission system for suspected abnormal hot spots, and issue an optical path warning signal when the judgment threshold is met.

2. The optical path fault prediction method based on machine learning according to claim 1 is characterized in that: The support vector machine algorithm is used to preliminarily screen the multi-channel real-time power data and temperature data in the target optical fiber transmission system to determine whether there are abnormal fluctuations that may trigger a fuse, including: Synchronously collect multiple channels of real-time power data and temperature data from the target optical fiber transmission system, wherein the multiple channels of real-time power data include the optical power measurement value of each transmission channel in the optical fiber, and the temperature data include the local temperature measurement value at the optical fiber transmission node; Performing data preprocessing on the collected multi-channel real-time power data and temperature data to generate cleaned multi-channel real-time power data and temperature data; A multi-dimensional feature space is constructed based on the cleaned multi-channel real-time power data and temperature data, and the multi-channel real-time power data and temperature data are used as inputs of feature vectors; The constructed multi-dimensional feature space is classified using a pre-trained support vector machine model to determine whether there is abnormal fluctuation. The support vector machine model is trained based on historical fiber fuse precursor data. During the training process, the classification boundary is optimized by setting abnormal sample labels. The result of the support vector machine model classification is output as determination information of whether there is abnormal fluctuation that may trigger fuse in the target optical fiber transmission system.

3. The optical path fault prediction method based on machine learning according to claim 1, characterized in that: The high frequency energy concentration is calculated by the following formula: ;in, Indicates the high-frequency energy concentration, It represents the sum of the energy of the coefficients after two-dimensional wavelet transformation in the high frequency range. represents the threshold of high frequency components, Represents the total energy over the entire frequency range; The low frequency energy dispersion is calculated by the following formula: ;in, represents the low-frequency energy diffusion, It represents the sum of the energy of the coefficients after two-dimensional wavelet transform in the low-frequency range.

4. The optical path fault prediction method based on machine learning according to claim 1, characterized in that: The calculation formula of spectrum characteristic index is: ;in, Represents the spectral characteristic index, which is defined as the energy integral within a predetermined frequency range. and Respectively represent the upper and lower limits of the frequency range, Represents the energy density in the frequency domain.

5. The optical path fault prediction method based on machine learning according to claim 1, characterized in that: The dynamic change characteristics of the non-uniform absorption pattern and the main features of the multiple reflection interference pattern are jointly extracted to obtain multi-dimensional fuse risk parameters, including: Obtain high-frequency energy concentration, low-frequency energy diffusion and spectrum characteristic index, perform standardization processing, and generate a multi-dimensional feature vector; Perform joint feature extraction on multi-dimensional feature vectors and use convolutional neural network algorithm to extract related features; A multidimensional fuse risk parameter for representing the optical fiber fuse risk is calculated based on the extracted correlation features.

6. The optical path fault prediction method based on machine learning according to claim 1, characterized in that: The judgment threshold is set according to the multi-dimensional fuse risk parameters, and the feature range overlapping with normal power fluctuations is screened out through adaptive rules, marking suspected abnormal hot spots, including: An initial determination threshold is set based on the distribution characteristics of the multi-dimensional fuse risk parameters, and the initial determination threshold is obtained by calculating the statistical distribution of the multi-dimensional fuse risk parameters; According to the historical data of absorption hot spots on the fiber end face and the dynamic change trend of multi-dimensional fuse risk parameters, the initial judgment threshold is optimized to obtain the judgment threshold; Reanalyze the distribution characteristics of absorption hot spots on the fiber end face and calculate the local absorption intensity and its spatial distribution range; Adaptive rules are applied to screen the absorption hotspot features of the reanalysis, and the feature ranges overlapping with normal power fluctuations are eliminated to mark suspected abnormal hotspot areas.

7. The optical path fault prediction method based on machine learning according to claim 1, characterized in that: Generate risk assessment results of impending fuse failure of the target optical fiber transmission system for suspected abnormal hot spots, and issue optical path warning signals when the judgment threshold is met, including: Obtain multi-dimensional fusing risk parameters of suspected abnormal hotspot areas, and generate a fusing risk assessment result of the target optical fiber transmission system based on the comparison result between the multi-dimensional fusing risk parameters and the judgment threshold; Output the fuse risk assessment results and optical path warning signals for subsequent optical path fault handling and protection strategy execution.

8. A machine learning-based optical path fault prediction system, used to implement a machine learning-based optical path fault prediction method according to any one of claims 1 to 7, characterized in that: It includes data screening module, frequency analysis module, scattering analysis module, feature extraction module, feature screening module and risk assessment module; Data screening module: Use the support vector machine algorithm to preliminarily screen the multi-channel real-time power data and temperature data in the target optical fiber transmission system to determine whether there are abnormal fluctuations that may trigger a fuse; Frequency analysis module: When there are abnormal fluctuations that may trigger fusing, the spatial frequency distribution of the absorption hotspots at the fiber end face is analyzed through two-dimensional wavelet transform to evaluate the dynamic change characteristics of the non-uniform absorption mode; Scattering analysis module: When there are abnormal fluctuations that may trigger a fuse, the interference intensity distribution of the fiber cladding multipath scattering is analyzed through the optical path tracing algorithm to identify the main features of the multiple reflection interference pattern; Feature extraction module: Joint feature extraction of the dynamic change characteristics of the non-uniform absorption pattern and the main features of the multiple reflection interference pattern to obtain multi-dimensional fuse risk parameters; Feature screening module: sets the judgment threshold according to the multi-dimensional fuse risk parameters, screens out the feature range overlapping with normal power fluctuations through adaptive rules, and marks suspected abnormal hot spots; Risk assessment module: Generates risk assessment results of impending fuse failure in the target optical fiber transmission system for suspected abnormal hot spots, and issues an optical path warning signal when the judgment threshold is met.

Images