A method and device for analyzing the health of an intelligent fiber distribution equipment

Abstract

The application provides a kind of intelligent optical fiber distribution equipment health degree analysis method and device, belong to optical fiber communication technical field, construct global LSTM model, and the initialized global model parameter is issued to each intelligent optical fiber distribution equipment, and local LSTM model training is carried out;Parameter alignment compensation and parameter aggregation strategy are carried out on global LSTM model using local training parameters obtained by each intelligent optical fiber distribution equipment, and updated global model parameters are obtained and issued to the local LSTM model of each intelligent optical fiber distribution equipment, to evaluate the current health status of each intelligent optical fiber distribution equipment.Thus, through the distributed federated learning framework, the local data distributed training of intelligent optical fiber distribution equipment and global model aggregation are realized, and then the data such as optical power loss and plug-in times are monitored and analyzed in real time, potential health risks are found in time, equipment damage caused by problem accumulation is avoided, and the adaptability and accuracy of equipment health monitoring are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of optical fiber communication, in particular to a health degree analysis method and device for intelligent optical fiber distribution equipment. BACKGROUND

[0002] With the explosive growth of business demands such as 5G and data centers, the scale of optical cable networks continues to expand, and the number of network nodes and the mileage of optical cables are growing exponentially, which puts forward urgent demands for intelligent operation and maintenance. Current optical fiber network operation and maintenance still mainly relies on manual operation, and fiber jump configuration and fault troubleshooting depend on manual on-site operation, which is low in efficiency and easily affected by human errors. With the continuous expansion of the scale of optical cable networks, intelligent optical fiber distribution equipment gradually replaces traditional manual mode, and functions such as automatic fiber jump and optical cable core quality monitoring are completed by mechanical arms to realize dynamic resource configuration and intelligent management of optical fiber networks. Such equipment can remotely complete core insertion and extraction, path switching and performance detection, which significantly improves the efficiency of operation and maintenance, but also puts forward higher requirements for accurate evaluation of the health status of the equipment.

[0003] Intelligent optical fiber distribution equipment has the characteristics of high-frequency plugging and strong adaptability to dynamic environment, and its health status is affected by multiple factors such as mechanical fatigue, optical signal attenuation, and environmental temperature and humidity. The existing health degree analysis methods for industrial equipment (such as motor vibration monitoring and bearing life prediction) are mostly based on offline data training models with fixed periods, which cannot adapt to the dynamic scene where instantaneous operation and long-term aging of optical fiber equipment coexist. In addition, the strong correlation between the optical signal characteristics (such as optical power loss) of optical fiber distribution equipment and mechanical movement (such as plugging force control) makes it difficult for traditional single-dimensional health evaluation models to accurately reflect the real state of the equipment. SUMMARY

[0004] Therefore, the purpose of the present application is to provide a health degree analysis method and device for intelligent optical fiber distribution equipment, which realizes distributed training of local data and aggregation of global models of intelligent optical fiber distribution equipment through a distributed federated learning framework, and improves the adaptability and accuracy of equipment health degree monitoring.

[0005] The health degree analysis method for intelligent optical fiber distribution equipment provided by the present application comprises the following steps:

[0006] Pretreat the equipment data collected by the intelligent optical fiber distribution equipment to construct time series data of different modalities;

[0007] A global LSTM model for predicting the health status of the intelligent optical fiber distribution equipment is constructed, and the initialized global model parameters are distributed to each intelligent optical fiber distribution equipment, so that each intelligent optical fiber distribution equipment trains a local LSTM model based on the time series data;

[0008] The local training parameters obtained by each intelligent fiber distribution equipment are used to perform parameter alignment compensation and parameter aggregation strategy on the global LSTM model, and updated global model parameters are obtained.

[0009] The updated global model parameters are loaded into the local LSTM model of each intelligent fiber distribution equipment, and the current health status of each intelligent fiber distribution equipment is evaluated.

[0010] In some embodiments, the method further comprises the following steps:

[0011] Monitoring whether the optical power loss and plug frequency fluctuation of each intelligent fiber distribution equipment are abnormal, if the optical power loss fluctuation or plug frequency fluctuation of a certain intelligent fiber distribution equipment exceeds the set threshold, retraining the local LSTM model based on the current and historical time series data within a period of time, and updating the global model parameters.

[0012] In some embodiments, the device data includes optical power loss, number of plug-in times of fiber connectors, lead screw running mileage and environmental data, the device data collected by the intelligent fiber distribution equipment is preprocessed, and time series data of different modalities are constructed, including the following steps:

[0013] Based on the sliding window mechanism, the number of plug-in times in each window is counted, and according to the preset threshold, it is divided into low-frequency plug-in scene and high-frequency plug-in scene; wherein, in the low-frequency plug-in scene, the optical power loss is processed by extreme value normalization; in the high-frequency plug-in scene, the optical power loss is processed by standard deviation normalization;

[0014] Based on the number of plug-in times, the lead screw running mileage and the time decay effect, the cumulative fatigue degree is calculated, which is used to measure the cumulative stress and loss of the equipment during use;

[0015] The timestamp of each plug-in operation is obtained, and according to the periodicity of the plug-in of the intelligent fiber distribution equipment, it is converted into first time encoding and second time encoding;

[0016] The normalized optical power loss, cumulative fatigue degree, number of plug-in times, first time encoding, second time encoding and environmental data are aligned by time step, forming a two-dimensional input tensor as time series data of different modalities.

[0017] In some embodiments, when each intelligent fiber distribution equipment trains the local LSTM model based on the time series data, the momentum stochastic gradient descent algorithm is used to adjust the local training parameters.

[0018] In some embodiments, the local training parameters obtained by each intelligent fiber distribution device perform parameter alignment compensation and parameter aggregation strategy on the global LSTM model, and obtain updated global model parameters, including the following steps:

[0019] Obtaining the local training parameters obtained by each intelligent fiber distribution device;

[0020] Calculating the first standard deviation of the optical power loss of each intelligent fiber distribution device within a set accumulation period, and the second standard deviation of the current optical power loss of all intelligent fiber distribution devices; and performing feature scaling compensation on the local training parameters based on the first standard deviation and the second standard deviation;

[0021] Calculating the number of optical power loss data samples, the historical average plug frequency, the cumulative fatigue degree of each intelligent fiber distribution device, and the total number of optical power loss data samples, the global average plug frequency, and the overall cumulative fatigue degree of the intelligent fiber distribution device; and performing aggregation on the local training parameters based on the set adjustment factor, the activity coefficient, the number of optical power loss data samples, the historical average plug frequency, the cumulative fatigue degree, the total number of optical power loss data samples, the global average plug frequency, and the overall cumulative fatigue degree, to obtain updated global model parameters.

[0022] In some embodiments, the current health status of each intelligent fiber distribution device is evaluated, including the following steps:

[0023] Inputting the constructed time series data after collection and preprocessing into the trained local LSTM model to obtain a health degree prediction value of the intelligent fiber distribution device;

[0024] Normalizing the health degree prediction value and mapping it to a set score interval, and according to the obtained score, representing the current health status of the intelligent fiber distribution device.

[0025] In some embodiments, the optical power loss and plug frequency fluctuation of each intelligent fiber distribution device are monitored to determine whether they are abnormal, including the following steps:

[0026] Calculating the third standard deviation of the optical power loss of each intelligent fiber distribution device within a statistical period, and if the third standard deviation exceeds a set optical power loss fluctuation threshold, determining that the optical power loss fluctuation is abnormal; wherein the statistical period is different for low-frequency plug-in scenarios and high-frequency plug-in scenarios;

[0027] If the number of plugs in the short-term time window and the long-term time window exceeds the set plug frequency fluctuation threshold, it is determined that the plug frequency fluctuation is abnormal; the plug frequency fluctuation threshold is set based on the average and standard deviation of the plug frequency in the short-term time window and the long-term time window.

[0028] In some embodiments, an intelligent optical distribution equipment health analysis device is also provided, and the device comprises:

[0029] a data collection preprocessing module configured to preprocess the equipment data collected by the intelligent optical distribution equipment, and to construct time series data of different modalities;

[0030] a model training module configured to construct a global LSTM model for predicting the health state of the intelligent optical distribution equipment, and to distribute the initialized global model parameters to each intelligent optical distribution equipment, so that each intelligent optical distribution equipment trains a local LSTM model based on the time series data;

[0031] a global model parameter aggregation module configured to perform parameter alignment compensation and parameter aggregation strategies on the global LSTM model using the local training parameters obtained by each intelligent optical distribution equipment, and to obtain updated global model parameters;

[0032] a health state evaluation module configured to load the updated global model parameters into the local LSTM model of each intelligent optical distribution equipment, and to evaluate the current health state of each intelligent optical distribution equipment.

[0033] In some embodiments, an electronic device is also provided, comprising a processor, a memory and a bus, the memory stores machine readable instructions executable by the processor, when the electronic device is running, the processor communicates with the memory through the bus, and the machine readable instructions are executed by the processor to perform the steps of the intelligent optical distribution equipment health analysis method described in any one of the above embodiments.

[0034] In some embodiments, a computer readable storage medium is also provided, and the computer readable storage medium stores a computer program, when the computer program is executed by a processor, the steps of the intelligent optical distribution equipment health analysis method described in any one of the above embodiments are performed.

[0035] The intelligent optical fiber distribution equipment health degree analysis method and device provided in the application preprocesses the equipment data collected by the intelligent optical fiber distribution equipment, constructs time series data of different modalities, constructs a global LSTM model for predicting the health state of the intelligent optical fiber distribution equipment, and distributes the initialized global model parameters to each intelligent optical fiber distribution equipment, so that each intelligent optical fiber distribution equipment trains a local LSTM model based on the time series data. The global LSTM model is executed by the local training parameters obtained by each intelligent optical fiber distribution equipment to perform parameter alignment compensation and parameter aggregation strategy, and updated global model parameters are obtained. The updated global model parameters are loaded into the local LSTM model of each intelligent optical fiber distribution equipment to evaluate the current health state of each intelligent optical fiber distribution equipment. Thus, through the distributed federated learning framework, the distributed training of local data of the intelligent optical fiber distribution equipment and the aggregation of the global model are realized, and the data such as optical power loss and plug-in frequency are monitored and analyzed in real time, potential health risks are found in time, and equipment damage caused by problem accumulation is avoided. Compared with the traditional manual operation mode, the adaptability and accuracy of the equipment health degree monitoring are improved. BRIEF DESCRIPTION OF DRAWINGS

[0036] In order to more clearly illustrate the technical solutions of the embodiments of the application, the following will briefly introduce the drawings needed to be used in the embodiments. It should be understood that the following drawings only show some embodiments of the application, and therefore should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can also be obtained without creative labor.

[0037] Figure 1 The flowchart of the intelligent optical fiber distribution equipment health degree analysis method described in the embodiments of the application is shown;

[0038] Figure 2 The flowchart of preprocessing the equipment data collected by the intelligent optical fiber distribution equipment and constructing time series data of different modalities described in the embodiments of the application is shown;

[0039] Figure 3 The flowchart of using the local training parameters obtained by each intelligent optical fiber distribution equipment to execute parameter alignment compensation and parameter aggregation strategy on the global LSTM model, and obtaining updated global model parameters described in the embodiments of the application is shown;

[0040] Figure 4 The flowchart of monitoring whether the optical power loss and plug-in frequency fluctuation of each intelligent optical fiber distribution equipment appear abnormal described in the embodiments of the application is shown;

[0041] Figure 5 The structural schematic diagram of the intelligent optical fiber distribution equipment health degree analysis device described in the embodiments of the application is shown;

[0042] Figure 6 A structural schematic diagram of the electronic device is shown. DETAILED DESCRIPTION

[0043] To make the objectives, technical solutions, and advantages of the embodiments of the present application clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present application with reference to the drawings in the embodiments of the present application. It should be understood that the drawings in the present application only serve the purpose of description and illustration, and are not used to limit the scope of protection of the present application. In addition, it should be understood that the schematic drawings are not drawn according to the actual proportions. The flowchart used in the present application shows the operations implemented according to some embodiments of the present application. It should be understood that the operations of the flowchart can not be implemented in sequence, and the steps without logical context relationship can be reversed in sequence or implemented simultaneously. In addition, one or more other operations can be added to the flowchart or one or more operations can be removed from the flowchart under the guidance of the content of the present application.

[0044] In addition, the described embodiments are only some of the embodiments of the present application, not all the embodiments. The components of the embodiments of the present application described and shown in the drawings herein can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the claimed present application, but only represents selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0045] It should be noted that the term “comprising” will be used in the embodiments of the present application to indicate the presence of the features declared thereafter, but does not exclude the addition of other features.

[0046] In view of the technical problems proposed in the background art, the present application provides a smart optical fiber distribution equipment health degree analysis method and device, electronic equipment and storage medium, which realizes distributed training of local data and aggregation of global model of smart optical fiber distribution equipment through a distributed federated learning framework, and improves the adaptability and accuracy of equipment health monitoring.

[0047] Referring to the drawings accompanying the Figure 1 The smart optical fiber distribution equipment health degree analysis method provided by the present application includes the following steps:

[0048] S1, pre-processing the equipment data collected by the smart optical fiber distribution equipment, and constructing time series data of different modalities;

[0049] S2, a global LSTM model for predicting the health status of the intelligent fiber distribution equipment is constructed, and the initialized global model parameters are distributed to each intelligent fiber distribution equipment, so that each intelligent fiber distribution equipment trains a local LSTM model based on the time series data;

[0050] S3, the global LSTM model is executed with the local training parameters obtained by each intelligent fiber distribution equipment to perform parameter alignment compensation and parameter aggregation strategy, and updated global model parameters are obtained;

[0051] S4, the updated global model parameters are loaded into the local LSTM model of each intelligent fiber distribution equipment, and the current health status of each intelligent fiber distribution equipment is evaluated.

[0052] In step S1, the device data collected by the intelligent fiber distribution equipment includes optical power loss of fiber channel, plug-in times of fiber connector, temperature, humidity, vibration intensity (time series data), screw running mileage and other optical, mechanical and environmental data. Among them, the parameter specification is attached Figure 2 The device data collected by the intelligent fiber distribution equipment is preprocessed, and time series data of different modalities are constructed, including the following steps:

[0053] S101, based on the sliding window mechanism, the plug-in times in each window are counted, and according to the preset threshold value, the low frequency plug-in scene and the high frequency plug-in scene are divided; wherein, in the low frequency plug-in scene, the optical power loss is processed by extreme value normalization; in the high frequency plug-in scene, the optical power loss is processed by standard deviation normalization;

[0054] S102, based on the plug-in times, the screw running mileage and the time attenuation effect, the cumulative fatigue degree is calculated, which is used to measure the cumulative stress and loss of the equipment in the use process;

[0055] S103, the time stamp of each plug-in operation is obtained, and according to the periodicity of the intelligent fiber distribution equipment plug-in, it is converted into first time encoding and second time encoding;

[0056] S104, the normalized optical power loss, cumulative fatigue degree, plug-in times, first time encoding, second time encoding and environmental data are aligned according to time step, forming a two-dimensional input tensor as time series data of different modalities.

[0057] In step S101, the optical power loss is dynamically normalized by using the plug frequency segmentation method. This is because, in the low-frequency plug scenario, the plug operation is less, the optical power loss data is relatively stable, and the noise proportion is high; in the high-frequency plug scenario, the plug is frequent, and the optical power loss fluctuates violently (such as a sudden increase in loss at the moment of plug). Therefore, the optical power loss data is divided into multiple subsets according to the plug frequency interval (such as low and high), and the normalization parameters (mean, standard deviation or extreme value) are calculated independently for each subset, so as to adapt to the difference in data distribution under different frequencies. Through plug frequency segmentation, "high-frequency amplification and low-frequency smoothing" is realized, the fluctuation characteristics of the optical power loss are directly related to the plug frequency, and the response ability of the model to extreme scenarios is improved.

[0058] In a specific implementation, a short-term time window Tshort (in days) can be defined by using a sliding window mechanism, and the number of plugs F in each window is counted. Since there are low-frequency plug and high-frequency plug scenarios, the number of plugs in the long-term time window Tlong is counted on the basis of counting the number of plugs in the time window. Tlong is a plurality of continuous time windows Tshort. The average plug frequency of the Tlong time window is evaluated by using an exponentially weighted moving average (EWMA).

[0059] F AVG (t) = a Ft + (1-a) F AVG (t-1)

[0060] Wherein, a is a forgetting factor, the greater a represents the smaller influence of historical data. Ft represents the number of plugs in the t time window; F AVG (t-1) represents the historical average number of plugs before the t time window. In an embodiment, the average plug frequency of the Tlong time window is divided into low frequency and high frequency by using a preset threshold value.

[0061] In the low-frequency plug scenario, the optical power loss is normalized by using the extreme value.

[0062] The maximum and minimum values in the Tlong time window are taken to calculate the normalized optical power loss data:

[0063]

[0064] Wherein, P is the optical power loss data collected by a certain intelligent optical fiber distribution device, P norm is the normalized data calculated, P max is the maximum value of P in the Tlong time window, and P minTlong is the minimum value of P in the time window, S is a scaling factor (range: 0.7-0.9, default: 0.8), and O is an offset (default 0.1). The purpose of setting S and O is to compress the normalized range to the middle interval and ensure that the minimum value is not zero, ensuring the physical meaning of the numerical value.

[0065] In the high-frequency plugging scenario, the optical power loss is normalized by standard deviation.

[0066]

[0067] wherein P avg is the mean optical power loss in the window, P σ is the standard deviation of the optical power loss in the window.

[0068] In step S102, the accumulated fatigue degree is an index for measuring the cumulative stress and loss of the device during long-term use. The intelligent fiber distribution device measures the degree of optical and mechanical wear and aging, i.e., the accumulated fatigue degree AEF, through the number of plugging times, the running mileage of the lead screw, and the effect of time attenuation.

[0069]

[0070] wherein, is the plugging dynamic attenuation, β is the basic damage amount of single plugging, usually 0.15; γ is the overload force multiple, usually 1.2, which takes effect when Fk>0.8Fmax; λ is the stress relaxation coefficient, usually 0.02 / hour according to the material, and the optical fiber joint is usually 0.02 / hour;

[0071] is the lead screw mileage dynamic wear, μ is the basic wear rate, usually 0.003 for stainless steel guide rail; α is the nonlinear wear acceleration index, usually 2.1; v crit The speed threshold is the value of the lead screw movement speed that significantly affects the wear of the device during the operation of the intelligent fiber distribution device. The value is determined according to the mechanical properties and design specifications of the device.

[0072] In step S103, the plug-in timestamp records the specific time of each plug-in operation. It can provide information about the time regularity, periodicity of the plug-in operation, convert the timestamp into a numerical way suitable for model processing, and serve as an input data set for collection. The plug-in timestamp is converted into a time interval, i.e. the time difference (hours) between the current time and the last plug-in. According to the periodic characteristics of the intelligent fiber distribution equipment plug-in, the monthly and daily periodicity of the plug-in operation is selected. Among them, time encoding 1 represents 30-day periodicity: convert the timestamp into the number of hours within 30 days (0-720), and embed it through sine / cosine coding: sin(2πt / 720), cos(2πt / 720);

[0073] Time encoding 2 represents daily periodicity: convert the timestamp into the number of hours within a day (0-23), and similarly, perform sine / cosine coding: sin(2πt / 24), cos(2πt / 24).

[0074] In step S104, the time series data of different modalities are aligned by time step to form a two-dimensional input tensor:

[0075] X t =[optical power, AEF, number of plug-ins, time encoding 1, time encoding 2, temperature, humidity]

[0076] Considering that the plug-in of the fiber connector has a great influence on the performance and health of the intelligent fiber distribution equipment, the number of plug-ins of the fiber connector at each time step is added as a new feature to Xt. The Min-Max normalization method can be used to convert the number of plug-ins n to

[0077] In step S2, a global long short-term memory network global LSTM model for predicting the health status of the intelligent fiber distribution equipment is first constructed. This model effectively processes time series data, captures the change rule of the connector plug-in frequency, cumulative fatigue, etc. over time, and completes model initialization. After that, the global model parameters such as LSTM layer weight, fully connected layer weight, LSTM layer bias, and fully connected layer bias are issued to each intelligent fiber distribution equipment. Each intelligent fiber distribution equipment receives the global model parameters, combines the locally collected and pre-processed time series data (optical power, cumulative fatigue, plug-in frequency, temperature, humidity), and inputs them into the local LSTM model of the intelligent fiber distribution equipment for training. The mean square error loss function MSE is used to calculate the mean square error between the predicted value and the true value, to evaluate the accuracy of the model prediction.

[0078]

[0079] Wherein, L represents the loss function value, used to measure the difference between the model prediction value and the true value, the smaller the loss function value, the closer the model prediction result to the true value, the better the performance of the model. N represents the total number of time series data samples participating in the model training, including optical power loss, cumulative fatigue degree, plug frequency and other information, the more the sample quantity, the more comprehensive the model learns the characteristics, the stronger the generalization ability of the model training result; X i represents the real health degree of the intelligent optical fiber distribution robot marked by the i th sample; represents the health degree prediction value of the intelligent optical fiber distribution robot obtained by the model analyzing the i th sample, that is, the output result obtained by the model after forward propagation calculation according to the current set parameters, the input sample data X i . In the local LSTM model training process, by continuously adjusting the model parameters, X i is as close as possible, so as to reduce the value of loss function L.

[0080] Further, in the intelligent optical fiber distribution equipment health state prediction model training process, considering the imbalance of different intelligent optical fiber distribution equipment data quantity, the processing efficiency of intelligent optical fiber distribution equipment limited by computing power, momentum stochastic gradient descent algorithm (Momentum SGD) is used to adjust the model parameters to obtain the smallest L value. The parameter adjustment process is as follows:

[0081] θ t = θ t-1 -v t

[0082] Wherein, θ t represents the parameter value of the model in the t th iteration, that is, the latest parameter value obtained after this update. θ t-1 represents the parameter value of the model in the t-1 th iteration, that is, the parameter value at the end of the last iteration. V t represents the momentum in the t th iteration, representing the cumulative effect of the gradient update direction; v t is calculated by the following formula:

[0083]

[0084] Wherein, t represents the current iteration step, indicating the current iteration number, the model will be iterated for many times, and the model parameters will be updated each time; μ is the momentum coefficient, the value range is between [0, 1], usually taking 0.9. μ determines the influence degree of the previous momentum on the current momentum. The greater μ is, the greater the influence of the past gradient information on the current update, and the model can maintain the previous motion direction in the update process; the smaller μ is, the relatively greater the influence of the current gradient; v t-1is the momentum at the t-1th iteration, representing the momentum value calculated in the last iteration. η represents the learning rate, which controls the step size of each parameter update. If the learning rate is too large, the model may oscillate around the optimal solution or even diverge; if the learning rate is too small, the convergence speed of the model will become very slow. represents the gradient of the loss function L with respect to the parameter θ at the t-1th iteration. The gradient represents the rate of change of the loss function at the current parameter value, and its direction points to the direction in which the loss function increases most rapidly, while the negative gradient direction is the direction in which the loss function decreases most rapidly.

[0085] After the intelligent fiber distribution equipment completes the local LSTM model parameter optimization, the updated LSTM layer weight, full connection layer weight, LSTM layer bias and full connection layer bias parameter are fed back to the global model for parameter alignment compensation and parameter aggregation strategy of the global model, and the model is iteratively updated.

[0086] In step S3, referring to the description accompanying drawings Figure 3 , the local training parameters obtained by each intelligent fiber distribution equipment are used to perform parameter alignment compensation and parameter aggregation strategy on the global LSTM model, and the updated global model parameters are obtained, including the following steps:

[0087] S301, obtaining local training parameters obtained by each intelligent fiber distribution equipment;

[0088] S302, calculating the first standard deviation of the optical power loss of each intelligent fiber distribution equipment in a set accumulation period, and the second standard deviation of the current optical power loss of all intelligent fiber distribution equipment; and performing feature scaling compensation on the local training parameters based on the first standard deviation and the second standard deviation;

[0089] S303, calculating the optical power loss data sample number, historical average plug frequency, cumulative fatigue degree of each intelligent fiber distribution equipment, and the sum of the optical power loss data sample number, global average plug frequency, and overall cumulative fatigue degree of the intelligent fiber distribution equipment; and performing aggregation on the local training parameters based on the set adjustment factor, activity coefficient, optical power loss data sample number, historical average plug frequency, cumulative fatigue degree, sum of optical power loss data sample number, global average plug frequency, and overall cumulative fatigue degree, to obtain the updated global model parameters.

[0090] In step S301, after the global model receives the parameters uploaded by the local training of the intelligent fiber distribution equipment, the parameter alignment compensation and parameter aggregation strategy in steps S302 and S303 are performed.

[0091] Specifically, in step S302, due to the large difference in optical power normalization parameters of low-frequency plug and high-frequency plug intelligent optical fiber distribution equipment, before federated aggregation, the local standard deviation and global standard deviation accumulated in the period are used for feature scaling compensation:

[0092]

[0093] wherein, is the original model parameter of the i-th intelligent optical fiber distribution equipment after local LSTM model training, which is obtained by training based on the local data of equipment i, and reflects the running characteristics of equipment i itself; is the local optical power loss standard deviation of the i-th intelligent optical fiber distribution equipment in the set accumulation period, which reflects the dispersion degree of the local optical power loss data of equipment i, and is compared with the global standard deviation to determine the difference degree of the data characteristics of equipment i and the global data characteristics; σ hlobal is the standard deviation of the current optical power loss data of all intelligent optical fiber distribution equipment, which represents the dispersion degree of the global data, and is used to unify the feature scales of different intelligent optical fiber distribution equipment. By calculating the standard deviation of the optical power loss data of all equipment, a global standard can be obtained to measure the fluctuation of different equipment data and avoid the influence of low-frequency plug intelligent optical fiber distribution equipment on the model; is the local model parameter of the i-th equipment after feature scaling compensation. This parameter not only considers the local optical power loss data characteristics of equipment i, but also adjusts the parameter by comparing it with the global standard deviation, so that it is more consistent with the requirements of the global model parameter.

[0094] In step S303, the plug frequency reflects the difference in equipment usage intensity. Different intelligent optical fiber distribution equipment has different plug frequencies, and the plug frequency has an impact on the optical power loss. When the plug frequency is low, the optical power loss data is relatively stable but the noise ratio is high; when the plug frequency is high, the optical power loss will surge in the plug moment and fluctuate violently. Considering the weight influence of low-frequency plug equipment and high-frequency plug equipment during parameter aggregation can more accurately learn the relationship between optical power loss and plug operation, improve the response ability of the model to the health status of the equipment in the extreme plug scenario, and ensure that the model output is more consistent with the actual operation of the equipment.

[0095] The accumulated fatigue degree measures the long-term health loss of the equipment. In the long-term operation of the intelligent optical fiber distribution equipment, the optical and mechanical parts are constantly worn and aged. The accumulated fatigue degree comprehensively considers the plug frequency, lead screw running mileage and other factors, and combines the time decay effect to measure the cumulative stress and loss of the equipment. The introduction of the accumulated fatigue degree enables the model to comprehensively consider the long-term running health risk accumulated by the equipment, avoids focusing only on the current plug operation, and supplements information from the long-term aging dimension of the equipment, which cooperates with short-term performance indicators such as optical power loss.

[0096]

[0097] wherein, represents the global model parameters updated after parameter aggregation, and the global model parameters are obtained by weighted summation of each local model parameter N i is the number of local optical power loss data samples of the ith intelligent fiber distribution device; N total is the sum of the number of local optical power loss data samples of all intelligent fiber distribution devices; F i is the historical average plug frequency (times / hour) of the ith intelligent fiber distribution device; is the global average plug frequency, which measures the plug frequency level of the overall device; AEF i is the cumulative fatigue degree of the ith intelligent fiber distribution device; represents the average cumulative fatigue degree of all intelligent fiber distribution devices, which measures the fatigue level of the overall device; α is an adjustment factor, and an increase in α enhances the weight of the intelligent fiber distribution device with high-frequency plug; α is usually 0.5; β i is the activity coefficient, which is determined according to the number of plugs in the last 30 days i (for example, β i = 1 + the number of plugs in the last 30 days), so as to ensure that the intelligent fiber distribution devices with different plug frequencies contribute more reasonably to the calculation results of the global model parameters.

[0098] The data volume, plug frequency and activity difference of different devices are considered by a weighted manner, the parameters of the intelligent fiber distribution device with more plug times have a larger weight in the global model aggregation, and the characteristics of the intelligent fiber distribution device with low-frequency plug are not ignored, so that the global model fuses the information of each local model more reasonably, and the adaptability and prediction accuracy for different device health states are improved.

[0099] After the global model completes double-weight aggregation and parameter alignment, the updated model parameters are sent to each intelligent fiber distribution device.

[0100] In step S4, after the global model completes double-weight aggregation and parameter alignment, the updated model parameters are sent to each intelligent fiber distribution device. Each intelligent fiber distribution device collects fiber-related data at the current time and in the past period of time (such as the past 24 hours), including optical power loss, plug times, temperature, humidity, etc. These data are sorted into time series data in the same format as during training according to time steps, an input tensor X is constructed, the prepared input tensor X is input into the trained local LSTM model, and the model outputs the device health degree prediction value.

[0101] Further, the health degree prediction value is normalized and mapped to a 0-100 score interval, denoted as Hscore, and the higher the score represents the better current health status of the intelligent optical fiber distribution equipment; the closer to 0, the worse the health status.

[0102] It can be seen that the intelligent optical fiber distribution equipment health degree analysis method provided by the application solves the problems of difficult collaborative optimization of different intelligent optical fiber distribution equipment models and inability to fully utilize overall data characteristics based on a federated learning framework. Through distributed modeling and global model aggregation, each device can share the training results of other devices to improve model performance and adaptability. This enables the model to more comprehensively capture the operation rules of each intelligent optical fiber distribution equipment in the network, accurately assess the health status of the equipment, meet the operation and maintenance needs of different intelligent optical fiber distribution equipment in a large-scale optical fiber network, and improve the overall operation and maintenance efficiency of the optical fiber network.

[0103] And the plug-in event label and the fatigue accumulation counter are embedded in the LSTM model. This way solves the problem of being difficult to capture the combined effects of instantaneous operation and long-term aging of intelligent optical fiber distribution equipment, breaks through the limitation of traditional models that can only process single type data or static data, and can dynamically and comprehensively model the device state, providing a guarantee for accurately assessing the health degree of the equipment, and then discovering potential problems in the equipment in a timely manner, ensuring stable operation of the intelligent optical fiber distribution equipment and enhancing the reliability of the optical fiber network.

[0104] In addition, when aggregating parameters, double weight distribution is performed considering the plug-in frequency and the cumulative fatigue degree, and an event alignment compensation mechanism is simultaneously executed. This solves the generalization problem of multi-device heterogeneous data, improves the limitation of traditional federated aggregation strategies that are difficult to handle data differences between devices, ensures that the global model reasonably fuses local model information of each intelligent optical fiber distribution equipment, enhances the adaptability and prediction accuracy of the health status of different equipment, and enables the model to be better applied to actual complex intelligent optical fiber distribution equipment scenarios.

[0105] In other embodiments, the intelligent optical fiber distribution equipment health degree analysis method provided by the application further includes the steps of:

[0106] S5, monitoring whether the optical power loss and plug-in frequency fluctuation of each intelligent optical fiber distribution equipment are abnormal, if the optical power loss fluctuation or plug-in frequency fluctuation of a certain intelligent optical fiber distribution equipment exceeds a set threshold, retraining the local LSTM model based on the time series data in the current and historical period and updating the global model parameters.

[0107] Specifically, referring to the drawings attached to the specification Figure 4 The monitoring of whether the optical power loss and plug-in frequency fluctuation of each intelligent optical fiber distribution equipment are abnormal includes the following steps:

[0108] S501, calculate the third standard deviation of the optical power loss of each intelligent optical fiber distribution device in a statistical period, and if the third standard deviation exceeds a set optical power loss fluctuation threshold, determine that the optical power loss fluctuation is abnormal; wherein the statistical period set for the low-frequency plug-in scene and the high-frequency plug-in scene is different;

[0109] S502, if the number of plugs in the short-term time window and the long-term time window exceeds the set plug-in frequency fluctuation threshold, determine that the plug-in frequency fluctuation is abnormal; the plug-in frequency fluctuation threshold is set based on the average value and standard deviation of the plug-in frequency in the short-term time window and the long-term time window.

[0110] In step S501, the optical power loss fluctuation monitoring and abnormality analysis are mainly performed. The optical power loss is a key indicator for measuring the performance of optical fiber transmission, and the fluctuation of optical power loss will affect the communication quality of the intelligent optical fiber distribution device providing optical cable core connection. In actual use, the optical power loss data is continuously collected, and the historical average fluctuation range of the optical power loss is calculated, denoted as ±σ, the optical power loss fluctuation ΔP is calculated for low-frequency and high-frequency scenes respectively, and the abnormal fluctuation of the optical power loss is determined.

[0111] In the low-frequency plug-in scene, a longer time period (such as a time period containing multiple T long ) is selected to calculate the optical power loss fluctuation; in the high-frequency plug-in scene, each plug-in operation is taken as an analysis unit, and the optical power loss value after each plug-in is recorded, and the average value of the optical power loss fluctuation in a period of time (such as a T long ) is calculated. The optical power loss fluctuation in the low-frequency scene and the average value of the optical power loss fluctuation in the high-frequency scene are uniformly denoted as ΔP. The standard deviation σ of the optical power loss in the statistical period is calculated, and if |ΔP|>kσ (k is a fluctuation abnormality coefficient), it is considered that the optical power loss fluctuation is abnormal.

[0112] In step S502, the plug-in frequency fluctuation monitoring and abnormality analysis are mainly performed. Specifically, during the operation of the intelligent optical fiber distribution device, the number of plugs of the intelligent optical fiber distribution device at the current time and in a period of time (the previous 24 hours) is continuously monitored and calculated, and the judgment results of the short-term and long-term time windows are considered to more accurately determine the plug-in frequency fluctuation.

[0113] In the short-term time window, the average value of the number of plugs F in a large number of historical T short is calculated to calculate the plug-in fluctuation; in the long-term time window, the average plug-in frequency F long in multiple consecutive T avg time windows is calculated to calculate the plug-in fluctuation. The average value of the short-term time window plug-in frequency F and the standard deviation σ short, long-term time window plug frequency average value and standard deviation σ long Short-term high-frequency threshold Long-term high-frequency threshold

[0114] In the short-term time window, when a certain T short , the number of insertions and extractions F short-high , it is determined that the insertion and extraction frequency in the short-term time window is abnormally fluctuating; in the long-term time window, when F avg >F long-high , it is determined that the insertion and extraction frequency in the long-term time window is abnormally fluctuating. When abnormal fluctuations occur in both the short-term and the long-term, or when abnormal fluctuations frequently occur in the short-term and the long-term fluctuation trend also deviates from the normal range, it is determined that the device insertion and extraction frequency is seriously abnormal.

[0115] When the optical power loss or the insertion and extraction frequency reaches the threshold value, the intelligent fiber distribution device starts model parameter updating. The intelligent fiber distribution device that has occurred abnormally inputs the optical power loss and the insertion and extraction frequency data in the current moment and in a period of time (such as the past 24 hours) into the local LSTM model, re-computes the optimized local model parameters, and uploads them to the global model. The global model obtains other local model parameters, executes the parameter aggregation strategy and parameter alignment compensation, and distributes the updated model parameters to all intelligent fiber distribution devices.

[0116] Based on the same inventive concept, the embodiments of the present application also provide an intelligent fiber distribution device health degree analysis device. Since the principle of solving problems in the device of the embodiments of the present application is similar to the above-mentioned intelligent fiber distribution device health degree analysis method, the implementation of the device can be referred to the implementation of the method, and the repeated parts will not be described here.

[0117] As shown in the accompanying drawings of the specification, Figure 5 the embodiments of the present application also provide an intelligent fiber distribution device health degree analysis device, which comprises:

[0118] The data acquisition and preprocessing module 501 is configured to preprocess the device data collected by the intelligent fiber distribution device, and construct time series data of different modalities.

[0119] The model training module 502 is configured to construct a global LSTM model for predicting the health state of the intelligent fiber distribution device, and distribute the initialized global model parameters to each intelligent fiber distribution device, so that each intelligent fiber distribution device trains a local LSTM model based on the time series data.

[0120] The global model parameter aggregation module 503 is configured to perform parameter alignment compensation and parameter aggregation strategies on the global LSTM model by using the local training parameters obtained by the respective intelligent fiber distribution equipment, and obtain updated global model parameters.

[0121] The health status evaluation module 504 is configured to load the updated global model parameters into the local LSTM model of each intelligent fiber distribution equipment, and evaluate the current health status of each intelligent fiber distribution equipment.

[0122] In an embodiment, the device data includes optical power loss, plug-in times of fiber connectors, lead screw running mileage, and environmental data. The device data collected by the intelligent fiber distribution equipment is preprocessed by the data collection and preprocessing module 501, and time series data of different modalities are constructed, including: based on a sliding window mechanism, the plug-in times in each window are counted, and according to a preset threshold value, the plug-in times are divided into a low-frequency plug-in scene and a high-frequency plug-in scene; wherein, in the low-frequency plug-in scene, the optical power loss is subjected to extreme value normalization processing; in the high-frequency plug-in scene, the optical power loss is subjected to standard deviation normalization processing; the cumulative fatigue degree is calculated based on the plug-in times, the lead screw running mileage, and the time decay effect, for measuring the cumulative stress and loss of the device in the use process; the timestamp of each plug-in operation is obtained, and is converted into first time encoding and second time encoding according to the periodicity of the plug-in of the intelligent fiber distribution equipment; the normalized optical power loss, the cumulative fatigue degree, the plug-in times, the first time encoding, the second time encoding, and the environmental data are aligned by time steps, forming a two-dimensional input tensor as time series data of different modalities.

[0123] In an embodiment, the global model parameter aggregation module 503 performs parameter alignment compensation and parameter aggregation strategies on the global LSTM model by using the local training parameters obtained by the respective intelligent fiber distribution equipment, and obtains updated global model parameters, including: obtaining the local training parameters obtained by the respective intelligent fiber distribution equipment; calculating the first standard deviation of the optical power loss of each intelligent fiber distribution equipment in a set cumulative period, and the second standard deviation of the current optical power loss of all intelligent fiber distribution equipment; and performing feature scaling compensation on the local training parameters based on the first standard deviation and the second standard deviation; calculating the optical power loss data sample number of each intelligent fiber distribution equipment, the historical average plug-in frequency, the cumulative fatigue degree, and the sum of the optical power loss data sample number of the intelligent fiber distribution equipment, the global average plug-in frequency, and the overall cumulative fatigue degree; and performing aggregation on the local training parameters based on the set adjustment factor, the activity coefficient, the optical power loss data sample number, the historical average plug-in frequency, the cumulative fatigue degree, the sum of the optical power loss data sample number, the global average plug-in frequency, and the overall cumulative fatigue degree, to obtain updated global model parameters.

[0124] In an embodiment, the health state evaluation module 504 evaluates the current health state of each intelligent fiber distribution device, including: inputting the constructed time series data after collection and preprocessing into the trained global LSTM model to obtain a health degree prediction value of the intelligent fiber distribution device at present; performing normalization processing on the health degree prediction value and mapping it to a set score interval, and according to the obtained score, representing the current health state of the intelligent fiber distribution device.

[0125] In an embodiment, the device further comprises:

[0126] The updating module is configured to monitor whether the optical power loss and the plug frequency fluctuation of each intelligent fiber distribution device are abnormal, and if the optical power loss fluctuation or the plug frequency fluctuation of a certain intelligent fiber distribution device exceeds a set threshold, retrain the local LSTM model based on the time series data in the current and historical period of time, and update the global model parameters. Wherein, a third standard deviation of the optical power loss of each intelligent fiber distribution device in a statistical period is calculated, and if the third standard deviation exceeds a set optical power loss fluctuation threshold, it is determined that the optical power loss fluctuation is abnormal; the statistical period is different for low-frequency plug-in and high-frequency plug-in scenarios; if the number of plugs in the short-term time window and the long-term time window exceeds the set plug frequency fluctuation threshold, it is determined that the plug frequency fluctuation is abnormal; the plug frequency fluctuation threshold is set based on the average value and the standard deviation of the plug frequency in the short-term time window and the long-term time window.

[0127] The intelligent fiber distribution device health degree analysis device described in the present application, through the data acquisition and processing module, pre-processes the device data collected by the intelligent fiber distribution device, and constructs time series data of different modalities; through the model training module, a global LSTM model for predicting the health state of the intelligent fiber distribution device is constructed, and the initialized global model parameters are distributed to each intelligent fiber distribution device, so that each intelligent fiber distribution device trains a local LSTM model based on the time series data; through the global model parameter aggregation module, the local training parameters obtained by each intelligent fiber distribution device are used to perform parameter alignment compensation and parameter aggregation strategy on the global LSTM model, and updated global model parameters are obtained; through the health state evaluation module, the updated global model parameters are loaded into the local LSTM model of each intelligent fiber distribution device to evaluate the current health state of each intelligent fiber distribution device. Thus, through the distributed federated learning framework, the distributed training of local data of the intelligent fiber distribution device and the global model aggregation are realized, and then the optical power loss, the number of plugs and other data are monitored and analyzed in real time, potential health risks are found in time, and equipment damage caused by problem accumulation is avoided. Compared with the traditional manual operation mode, the adaptability and accuracy of the equipment health degree monitoring are improved.

[0128] Based on the same concept of the present application, as shown in the description Figure 6 The electronic device 600 provided by the embodiment of the present application includes at least one processor 601, at least one network interface 604 or other user interface 603, a memory 605, and at least one communication bus 602. The communication bus 602 is used to realize the connection and communication between the components. The electronic device 600 can optionally include a user interface 603, including a display (for example, a touch screen, an LCD, a CRT, holographic imaging (Holographic), or a projector, etc.), a keyboard or a clicking device (for example, a mouse, a trackball, a touchpad, or a touch screen, etc.).

[0129] The memory 605 can include read-only memory and random access memory, and provide instructions and data for the processor 601. A part of the memory 605 can also include a non-volatile random access memory (NVRAM).

[0130] In some embodiments, the memory 605 stores the following elements, executable modules or data structures, or a subset of them, or an extended set of them:

[0131] The operating system 6051 includes various system programs, used to realize various basic services and process hardware-based tasks;

[0132] The application program module 6052 includes various application programs, such as a desktop (launcher), a media player (MediaPlayer), a browser (Browser), etc., used to realize various application services.

[0133] In the embodiment of the present application, the processor 601 is used to execute the steps of the method for analyzing the health degree of the intelligent optical fiber distribution equipment by calling the programs or instructions stored in the memory 605.

[0134] The present application also provides a computer readable storage medium, which stores a computer program. When the computer program is run by a processor, the steps of the method for analyzing the health degree of the intelligent optical fiber distribution equipment are executed.

[0135] Specifically, the storage medium can be a general storage medium, such as a mobile disk, a hard disk, etc. When the computer program on the storage medium is run, the distributed local data distributed training and global model aggregation of the intelligent distribution equipment can be realized through the distributed federated learning framework, and the adaptability and accuracy of the equipment health monitoring are improved.

[0136] In the embodiments of the present application, it should be understood that the disclosed apparatus and method can be implemented in other manners. The embodiments described above are merely exemplary, for example, the division of units is only a logical function division, and there can be another division manner in actual implementation; for example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the displayed or discussed coupling or direct coupling or communication connection between units can be indirect coupling or communication connection through some interfaces, and can be electrical, mechanical or other forms.

[0137] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they can be located in one place, or can be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the embodiments.

[0138] In addition, each functional unit in the embodiments of the present application can be integrated into a processing unit, or each unit can be a physically independent unit, or two or more units can be integrated into one unit.

[0139] If the functions are implemented in the form of software function units and sold or used as independent products, they can be stored in a computer readable storage medium. Based on this understanding, the technical solutions of the present application essentially or the part that contributes to the prior art or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium, and includes a number of 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 in the various embodiments of the present application. The foregoing storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), magnetic disk or optical disk, and various program codes that can store program codes.

[0140] Finally, it should be noted that the above examples are merely specific embodiments of the present application, and are used to illustrate the technical solutions of the present application, but not to limit the same. The protection scope of the present application is not limited thereto. Although the present application has been described in detail with reference to the foregoing examples, it should be understood by those skilled in the art that any person skilled in the art can make modifications or easily think of changes to the technical solutions recorded in the foregoing examples, or make equivalent replacements to some of the technical features, within the technical scope disclosed by the present application. The modifications, changes or replacements do not make the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application. All of them should be covered in 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.

Claims

1. A method for health analysis of intelligent fiber optic distribution equipment, characterized in that, The method includes the following steps: The equipment data collected by the intelligent fiber optic distribution equipment is preprocessed to construct time-series data for different modes. The equipment data includes optical power loss, fiber optic connector insertion / removal counts, lead screw mileage, and environmental data. The preprocessing of the equipment data to construct time-series data for different modes includes the following steps: counting the insertion / removal counts within each window based on a sliding window mechanism, and dividing the data into low-frequency and high-frequency insertion / removal scenarios according to preset threshold values; wherein, in the low-frequency insertion / removal scenario, extreme value normalization is applied to the optical power loss; in the high-frequency insertion / removal scenario... In frequent plugging and unplugging scenarios, the optical power loss is normalized using standard deviation. The cumulative fatigue is calculated based on the number of plugging and unplugging operations, the mileage of the lead screw, and the time decay effect, which is used to measure the cumulative pressure and loss borne by the equipment during use. The timestamp of each plugging and unplugging operation is obtained and converted into a first time code and a second time code according to the periodicity of the plugging and unplugging of the intelligent fiber optic distribution equipment. The normalized optical power loss, cumulative fatigue, number of plugging and unplugging operations, first time code, second time code, and environmental data are aligned by time step to form a two-dimensional input tensor, which serves as time series data for different modes. A global LSTM model for predicting the health status of intelligent fiber optic distribution equipment is constructed, and the initialized global model parameters are distributed to each intelligent fiber optic distribution equipment so that each intelligent fiber optic distribution equipment can train a local LSTM model based on the time series data. The following steps are used to perform parameter alignment compensation and parameter aggregation strategies on the global LSTM model using the local training parameters obtained from each intelligent fiber optic distribution device, and to obtain updated global model parameters: obtaining the local training parameters obtained from each intelligent fiber optic distribution device; calculating the first standard deviation of optical power loss of each intelligent fiber optic distribution device within a set cumulative period, and the second standard deviation of current optical power loss of all intelligent fiber optic distribution devices; and performing feature scaling compensation on the local training parameters based on the first and second standard deviations; calculating the number of optical power loss data samples, historical average insertion / removal frequency, and cumulative fatigue of each intelligent fiber optic distribution device, as well as the total number of optical power loss data samples, global average insertion / removal frequency, and overall cumulative fatigue of the intelligent fiber optic distribution devices; and aggregating the local training parameters based on a set adjustment factor, activity coefficient, number of optical power loss data samples, historical average insertion / removal frequency, cumulative fatigue, total number of optical power loss data samples, global average insertion / removal frequency, and overall cumulative fatigue to obtain updated global model parameters. The updated global model parameters are loaded into the local LSTM model of each smart fiber optic distribution device to evaluate the current health status of each smart fiber optic distribution device.

2. The method for health analysis of intelligent optical fiber distribution equipment according to claim 1, characterized in that, The method further includes the following steps: Monitor whether there are any abnormalities in the optical power loss and insertion / removal frequency fluctuations of each intelligent fiber optic distribution device. If the optical power loss fluctuation or insertion / removal frequency fluctuation of a certain intelligent fiber optic distribution device exceeds the set threshold, retrain the local LSTM model and update the global model parameters based on the current and historical time series data.

3. The health analysis method for intelligent optical fiber distribution equipment according to claim 1, characterized in that, in, When each intelligent fiber optic distribution device trains its local LSTM model based on the time series data, it uses the momentum stochastic gradient descent algorithm to adjust the local training parameters.

4. The method for health analysis of intelligent optical fiber distribution equipment according to claim 1, characterized in that, The assessment of the current health status of each intelligent fiber optic distribution device includes the following steps: The collected and preprocessed time series data is input into the trained local LSTM model to obtain the current health prediction value of the intelligent fiber optic distribution equipment. The predicted health values ​​are normalized and mapped to a set score range. The obtained scores represent the current health status of the intelligent fiber optic distribution equipment.

5. The method for health analysis of intelligent optical fiber distribution equipment according to claim 2, characterized in that, The monitoring of whether the optical power loss and insertion / removal frequency fluctuations of each intelligent fiber optic distribution device are abnormal includes the following steps: The third standard deviation of optical power loss of each intelligent fiber optic distribution device within the statistical period is calculated. If the third standard deviation exceeds the set optical power loss fluctuation threshold, it is determined that the optical power loss fluctuation is abnormal. The statistical period is set differently for low-frequency plugging and unplugging scenarios and high-frequency plugging and unplugging scenarios. If the number of insertions and removals exceeds the set insertion and removal frequency fluctuation threshold in both the short-term and long-term time windows, it is determined that the insertion and removal frequency fluctuation is abnormal; the insertion and removal frequency fluctuation threshold is set based on the average value and standard deviation of the insertion and removal frequency in the short-term and long-term time windows.

6. A health analysis device for intelligent fiber optic distribution equipment, characterized in that, The device includes: The data acquisition and preprocessing module is used to preprocess the equipment data collected by the intelligent fiber optic distribution equipment to construct time series data of different modes. The equipment data includes optical power loss, fiber optic connector insertion / removal counts, lead screw mileage, and environmental data. The preprocessing of the equipment data to construct time series data of different modes includes: counting the insertion / removal counts within each window based on a sliding window mechanism, and dividing the scenarios into low-frequency and high-frequency insertion / removal scenarios according to preset threshold values. In the low-frequency insertion / removal scenario, the optical power loss is normalized using extreme value normalization. In high-frequency insertion and removal scenarios, the optical power loss is normalized using standard deviation. Cumulative fatigue is calculated based on the number of insertions and removals, the lead screw mileage, and the time decay effect, used to measure the cumulative pressure and loss experienced by the equipment during use. The timestamp of each insertion and removal operation is obtained and converted into a first time code and a second time code according to the periodicity of the intelligent fiber optic distribution equipment's insertion and removal. The normalized optical power loss, cumulative fatigue, number of insertions and removals, first time code, second time code, and environmental data are aligned by time steps to form a two-dimensional input tensor, serving as time series data for different modes. The model training module is used to construct a global LSTM model for predicting the health status of intelligent fiber optic distribution equipment, and to distribute the initialized global model parameters to each intelligent fiber optic distribution equipment so that each intelligent fiber optic distribution equipment can perform local LSTM model training based on the time series data. The global model parameter aggregation module is used to perform parameter alignment compensation and parameter aggregation strategies on the global LSTM model using the local training parameters obtained from each intelligent fiber optic distribution device, and obtain updated global model parameters. This includes: acquiring the local training parameters obtained from each intelligent fiber optic distribution device; calculating the first standard deviation of optical power loss for each intelligent fiber optic distribution device within a set cumulative period, and the second standard deviation of current optical power loss for all intelligent fiber optic distribution devices; performing feature scaling compensation on the local training parameters based on the first and second standard deviations; calculating the number of optical power loss data samples, historical average insertion / removal frequency, and cumulative fatigue for each intelligent fiber optic distribution device, as well as the total number of optical power loss data samples, global average insertion / removal frequency, and overall cumulative fatigue for the intelligent fiber optic distribution devices; and aggregating the local training parameters based on a set adjustment factor, activity coefficient, number of optical power loss data samples, historical average insertion / removal frequency, cumulative fatigue, total number of optical power loss data samples, global average insertion / removal frequency, and overall cumulative fatigue to obtain updated global model parameters. The health status assessment module is used to load the updated global model parameters into the local LSTM model of each intelligent fiber optic distribution device to assess the current health status of each intelligent fiber optic distribution device.

7. An electronic device, characterized in that, include: The device includes a processor, a memory, and a bus. The memory stores machine-readable instructions executable by the processor. When the electronic device is running, the processor communicates with the memory via the bus. When the machine-readable instructions are executed by the processor, they perform the steps of a health analysis method for intelligent fiber optic distribution equipment as described in any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, performs the steps of the health analysis method for an intelligent fiber optic distribution device as described in any one of claims 1 to 5.

Images