Optical communication device performance monitoring method and system based on big data analysis

Abstract

The present application provides a method and system for monitoring the performance of an optical communication device based on big data analysis, which relates to the field of optical communication equipment management. The method includes: obtaining real-time operation data of an optical communication device; extracting the equipment operation features corresponding to the real-time operation data based on a feature extraction model; extracting the abnormal behavior pattern features corresponding to the equipment operation features based on an abnormal pattern analysis model; inputting the real-time operation data, the equipment operation features and the abnormal behavior pattern features into a fault identification model, so that the fault identification model outputs the corresponding target fault type; and outputting the equipment performance indicators of the optical communication device according to the real-time operation data when the first abnormal behavior pattern or the second abnormal behavior pattern is not detected. Thus, the big data analysis and machine learning model technologies are comprehensively utilized to identify potential equipment failures, which significantly improves the operation stability and maintenance management efficiency of the optical communication device.

Description

Technical Field

[0001] The present application relates to the technical field of optical communication equipment management, and in particular to a method and system for monitoring the performance of optical communication devices based on big data analysis. Background Art

[0002] Optical communication devices usually include optical transmitters and optical receivers, which are important components of modern communication networks. The stability and reliability of communication equipment performance will directly affect the communication quality.

[0003] At present, the monitoring method for optical communication devices is mainly to monitor and analyze physical parameter indicators, such as temperature and other parameters, and trigger an alarm when the device parameters exceed the alarm threshold. For example, when the device temperature is too high, it may mean that the device is overloaded or has a potential fault. However, monitoring only a small number of parameters cannot fully reflect the operating status of the device. In addition, due to the components or application environment of different devices, the alarm thresholds of different devices require a lot of experiments and debugging, which increases the difficulty of equipment management and debugging.

[0004] Based on the current equipment alarm mechanism, it can only monitor equipment failures that have occurred, but cannot predict failure risks, resulting in communication interruptions during equipment failure maintenance time, seriously affecting the user's communication service experience. In addition, after the equipment alarm, professionals are required to debug according to the alarm signal and manually identify the equipment failure, resulting in longer equipment maintenance time and costs.

[0005] In response to the above problems, the industry has not yet proposed a better technical solution. Summary of the invention

[0006] The present application provides a method, system, storage medium, computer program product and electronic device for monitoring the performance of an optical communication device based on big data analysis, which are used to at least solve the problems of incomplete equipment monitoring, insufficient fault warning capability and high maintenance cost of current optical communication devices.

[0007] In a first aspect, an embodiment of the present application provides a method for monitoring the performance of an optical communication device based on big data analysis, including: acquiring real-time operating data of an optical communication device; the real-time operating data includes multiple types of operating parameters: internal temperature of the device, ambient temperature and humidity, real-time current, real-time voltage and optical power; extracting the device operating characteristics corresponding to the real-time operating data based on a feature extraction model; the feature extraction model includes a principal component analysis module, a linear discriminant analysis module and an autoencoder, the principal component analysis module is used to perform dimensionality reduction processing on the real-time operating data, and then perform discriminant feature extraction through the linear discriminant analysis module, and then perform nonlinear feature learning through the autoencoder to obtain corresponding device operating characteristics; extracting abnormal behavior pattern characteristics corresponding to the device operating characteristics based on an abnormal pattern analysis model, specifically including: the abnormal pattern analysis model includes an association rule mining module and a cluster analysis module; the association rule mining module is used to perform association rule mining on the device operation feature data to identify whether there is a first abnormal behavior mode; the cluster analysis module is used to perform cluster analysis on the equipment operation characteristic data to identify whether there is a second abnormal behavior mode, and when the first abnormal behavior mode and / or the second abnormal behavior mode are detected, the abnormal behavior mode characteristics are determined according to the first abnormal behavior characteristics obtained by the association rule mining and the second abnormal behavior characteristics obtained by the cluster analysis; the real-time operation data, the equipment operation characteristics and the abnormal behavior mode characteristics are input into the fault identification model, so that the fault identification model outputs the corresponding target fault type; the target fault type includes any one of the following: laser aging fault, laser temperature drift fault, laser drive circuit fault, optical receiver sensitivity fault and optical amplifier gain abnormality; when the first abnormal behavior mode or the second abnormal behavior mode is not detected, the equipment performance index of the optical communication device is output according to the real-time operation data; the equipment performance index includes equipment operation stability, equipment energy efficiency ratio, signal quality ratio and power conversion efficiency.

[0008] In the second aspect, the embodiment of the present application provides a performance monitoring system for an optical communication device based on big data analysis, including: an operation data acquisition unit, used to acquire real-time operation data of the optical communication device; the real-time operation data includes multiple types of operation parameters: internal temperature of the equipment, ambient temperature and humidity, real-time current, real-time voltage and optical power; an operation feature extraction unit, used to extract the equipment operation characteristics corresponding to the real-time operation data based on a feature extraction model; the feature extraction model includes a principal component analysis module, a linear discriminant analysis module and an autoencoder, the principal component analysis module is used to perform dimensionality reduction processing on the real-time operation data, and then perform discriminant feature extraction through the linear discriminant analysis module, and then perform nonlinear feature learning through the autoencoder to obtain corresponding equipment operation characteristics; a feature pattern analysis unit, used to extract abnormal behavior pattern characteristics corresponding to the equipment operation characteristics based on an abnormal pattern analysis model, specifically including: the abnormal pattern analysis model includes an association rule mining module and a clustering analysis module; the association rule mining module is used to perform association rule mining on the equipment operation feature data to identify whether there is a first abnormal behavior pattern; the cluster analysis module is used to perform cluster analysis on the device operation characteristic data to identify whether there is a second abnormal behavior pattern, and when the first abnormal behavior pattern and / or the second abnormal behavior pattern are detected, determine the abnormal behavior pattern characteristics according to the first abnormal behavior characteristics obtained by the association rule mining and the second abnormal behavior characteristics obtained by the cluster analysis; a fault type analysis unit is used to input the real-time operation data, the device operation characteristics and the abnormal behavior pattern characteristics into a fault identification model, so that the fault identification model outputs the corresponding target fault type; the target fault type includes any one of the following: laser aging fault, laser temperature drift fault, laser drive circuit fault, optical receiver sensitivity fault and optical amplifier gain abnormality; a performance indicator monitoring unit is used to output the device performance indicator of the optical communication device according to the real-time operation data when the first abnormal behavior pattern or the second abnormal behavior pattern is not detected; the device performance indicator includes device operation stability, device energy efficiency ratio, signal quality ratio and power conversion efficiency.

[0009] According to a third aspect, an electronic device is provided, comprising: at least one processor, and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can perform the steps of the optical communication device performance monitoring method based on big data analysis of any embodiment of the present application.

[0010] In a fourth aspect, an embodiment of the present application provides a storage medium on which a computer program is stored, characterized in that when the program is executed by a processor, the steps of the optical communication device performance monitoring method based on big data analysis of any embodiment of the present application are implemented.

[0011] In a fifth aspect, an embodiment of the present application provides a computer program product, including a computer program / instruction, which, when executed by a processor, implements the steps of the optical communication device performance monitoring method based on big data analysis of any embodiment of the present application.

[0012] The optical communication device performance monitoring method based on big data analysis provided by the present application can produce at least the following technical effects:

[0013] (1) By acquiring the real-time operating data of the optical communication device, including various parameters such as the internal temperature of the device, ambient temperature and humidity, real-time current, real-time voltage and optical power, the operating status of the device can be monitored in all directions, providing more comprehensive and detailed monitoring, and more accurately reflecting the actual operating status of the device.

[0014] (2) By introducing a feature extraction model that combines principal component analysis (PCA), linear discriminant analysis (LDA) and autoencoder, the real-time operation data is reduced in dimension, discriminant features are extracted and nonlinear features are learned. In this way, the key information in the equipment operation data can be retained to the maximum extent, and redundancy and noise can be removed, thereby improving the accuracy and effect of feature extraction.

[0015] (3) By extracting the abnormal behavior pattern characteristics corresponding to the equipment operation characteristics through association rule mining and cluster analysis, multi-level identification of abnormal behavior is achieved. Association rule mining can discover the implicit relationship and potential patterns between equipment operation characteristics, while cluster analysis can classify according to the similarity of characteristics. This two-pronged approach makes abnormal pattern identification more accurate and reduces false alarm and missed alarm rates.

[0016] (4) The fault identification model processes real-time operation data, equipment operation characteristics, and abnormal behavior pattern characteristics, and outputs specific target fault types. Based on multi-dimensional data input, the fault identification model has higher accuracy and reliability, and can achieve early prediction of fault types, making it easier for managers to take maintenance measures in advance, thereby avoiding communication interruptions caused by sudden equipment failures.

[0017] (5) Through big data analysis technology, association rule mining and cluster analysis are used to conduct in-depth analysis of equipment operation characteristics, which can quickly and effectively identify whether there are abnormal behavior patterns. When an abnormal behavior pattern is identified, the fault identification model is called to predict and output the corresponding target fault type. There is no need for management personnel to repeatedly debug to locate the fault, which significantly reduces the time and cost of equipment debugging and maintenance. In addition, when no abnormal behavior pattern is detected, the system can output equipment performance indicators based on real-time operation data to improve equipment management efficiency.

[0018] Through this technical solution, through accurate fault identification and early warning mechanisms, and using machine learning model technology to predict potential faults in advance, maintenance can be carried out before faults occur to avoid communication interruptions, thereby improving the stability and reliability of communication services and significantly improving the operating stability and maintenance management efficiency of the equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0020] Figure 1 A flowchart showing an example of a method for monitoring performance of an optical communication device based on big data analysis according to an embodiment of the present application is shown;

[0021] Figure 2 A flowchart of an example of extracting abnormal behavior pattern features according to an embodiment of the present application is shown;

[0022] Figure 3 Shown according to Figure 1 An exemplary operation flow chart of step S110 in FIG.

[0023] Figure 4 A schematic diagram showing a structural connection of an example of an abnormal pattern analysis model according to an embodiment of the present application is shown;

[0024] Figure 5 An operation flow chart of an example of identifying a first abnormal behavior pattern by an association rule mining module according to an embodiment of the present application is shown;

[0025] Figure 6 An operation flow chart of an example of identifying a second abnormal behavior pattern by a cluster analysis module according to an embodiment of the present application is shown;

[0026] Figure 7An operation flow chart of an example of determining abnormal behavior pattern characteristics according to an embodiment of the present application is shown;

[0027] Figure 8 A schematic diagram of structural connection of an example of a fault identification model according to an embodiment of the present application is shown;

[0028] Fig. 9 A structural block diagram of an example of an optical communication device performance monitoring system based on big data analysis according to an embodiment of the present application is shown;

[0029] Fig.10 It is a schematic structural diagram of an embodiment of an electronic device of the present application. DETAILED DESCRIPTION

[0030] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0031] In the technical solution of this application, the collection, storage, use, processing, transmission, provision and disclosure of user personal information involved shall comply with the provisions of relevant laws and regulations and shall not violate public order and good morals.

[0032] Figure 1 A flowchart of an example of a method for monitoring the performance of an optical communication device based on big data analysis according to an embodiment of the present application is shown.

[0033] Regarding the executor of the method of the embodiment of the present application, it can be any controller or processor with computing or processing capabilities. Through automated data analysis and fault identification, early prediction and identification of faults can avoid sudden failures of equipment during operation, reducing maintenance costs and economic losses from communication interruptions.

[0034] In some examples, it can be an optical communication device monitoring platform, and can be integrated and configured in an electronic device or terminal through software, hardware, or a combination of software and hardware, and the type of terminal or electronic device can be diverse, such as a mobile phone, tablet computer, or desktop computer, etc.

[0035] like Figure 1 As shown, in step S110, real-time operation data of the optical communication device is acquired.

[0036] In some embodiments, multiple sensors are deployed at key locations of the optical communication device to collect real-time data such as the internal temperature of the device, ambient temperature and humidity, real-time current, real-time voltage, and optical power. Furthermore, a high-speed data acquisition system is used to ensure high-frequency acquisition and transmission of real-time data to avoid data delay or loss. The collected data is preprocessed, including denoising, filtering, and standardization, to ensure data accuracy and consistency.

[0037] Therefore, by acquiring various operating parameters of the optical communication device in real time, the operating status of the equipment can be fully reflected and captured in time, providing comprehensive data support for subsequent feature extraction and abnormal pattern analysis, which helps to timely discover potential faults or abnormal conditions.

[0038] In step S120, the equipment operation features corresponding to the real-time operation data are extracted based on the feature extraction model.

[0039] Here, the feature extraction model includes the principal component analysis (PCA) module, the linear discriminant analysis (LDA) module, and the autoencoder. Through PCA processing, the real-time operation data is reduced in dimension to reduce data redundancy, retain the main information, and improve the efficiency of data processing and the accuracy of feature extraction. In addition, through LDA processing, discriminant feature extraction is performed on the basis of the reduced-dimensional data to enhance the identifiability of the features and distinguish different equipment operation states. Furthermore, the autoencoder performs nonlinear learning on the features to further extract deep-level features and improve the depth and accuracy of feature extraction.

[0040] Therefore, PCA dimensionality reduction processing can reduce data redundancy and improve the efficiency of feature extraction. The combination of LDA and autoencoder enhances the recognition ability and depth of features, ensuring that the extracted features can accurately reflect the operating status of the equipment.

[0041] In step S130, abnormal behavior pattern features corresponding to the equipment operation features are extracted based on the abnormal pattern analysis model.

[0042] More specifically, the abnormal pattern analysis model includes an association rule mining module and a cluster analysis module. Figure 2 FIG. 1 is a flowchart showing an example of extracting abnormal behavior pattern features according to an embodiment of the present application. Figure 2As shown, in step S210, the association rule mining module performs association rule mining on the device operation feature data to identify whether there is a first abnormal behavior pattern. In step S220, the cluster analysis module performs cluster analysis on the device operation feature data to identify whether there is a second abnormal behavior pattern. In step S230, when the first abnormal behavior pattern and / or the second abnormal behavior pattern is detected, the abnormal behavior pattern feature is determined based on the first abnormal behavior feature obtained by the association rule mining and the second abnormal behavior feature obtained by the cluster analysis.

[0043] Here, whether the abnormal risk is discovered through association rule mining or through cluster analysis, as long as it is determined that the equipment has an abnormal operation risk, the features of the association rule mining can be combined with the features output by the cluster analysis to obtain the corresponding abnormal behavior pattern features.

[0044] Therefore, by combining association rule mining and cluster analysis, we can ensure comprehensive identification of different types of abnormal behaviors, improve the accuracy of abnormal behavior pattern identification, and reduce the risks of false alarms and missed alarms.

[0045] In step S141 , real-time operation data, equipment operation characteristics, and abnormal behavior pattern characteristics are input into a fault identification model so that the fault identification model outputs a corresponding target fault type.

[0046] In some embodiments, various advanced machine learning algorithms (such as random forest, support vector machine or deep neural network) can be used to build a fault identification model, and then a large amount of historical fault data is used to train the fault identification model, optimize the model parameters, and improve the recognition accuracy of the model. In this way, in practical applications, real-time operation data, equipment operation characteristics and abnormal behavior pattern characteristics are input into the fault identification model, and the corresponding target fault type is output. Here, the target fault type includes any one of the following: laser aging fault, laser temperature drift fault, laser drive circuit fault, optical receiver sensitivity fault and optical amplifier gain abnormality.

[0047] Therefore, through an efficient fault identification model, different types of equipment failures can be accurately identified and potential equipment problems can be warned in a timely manner.

[0048] On the other hand, in step S143 , when the first abnormal behavior pattern or the second abnormal behavior pattern is not detected, the device performance index of the optical communication apparatus is output according to the real-time operation data.

[0049] Here, the device performance index includes device operation stability, device energy efficiency ratio, signal quality ratio and power conversion efficiency. In some embodiments, when no abnormal behavior pattern is detected, the device operation stability, energy efficiency ratio, signal quality ratio and power conversion efficiency and other performance indicators are calculated based on real-time operation data.

[0050] In some embodiments, the device energy efficiency ratio (EER) can be determined by calculating the ratio between the optical power output and the electrical energy consumed by the device (the product of current and voltage). The device operation stability can be quantified by monitoring the fluctuations of the internal temperature, real-time current and real-time voltage of the device, and by statistical methods such as standard deviation and root mean square (RMS) value, so as to obtain the operation stability of the computing device within a certain period of time. The signal quality ratio (SQR) can be determined by calculating the ratio between the optical power and the environmental impact factor, and the environmental impact factor can be quantified by the ambient temperature and humidity and the internal temperature of the device. The power conversion efficiency can be determined by calculating the ratio between the electrical power input and the optical power output of the optical communication device.

[0051] In this embodiment, the device performance indicators are displayed through a visual interface to provide intuitive device status information to the operation and maintenance personnel. Thus, by updating multiple device performance indicators in real time, the operation and maintenance personnel are helped to comprehensively evaluate the operating status of the device.

[0052] Through the embodiments of the present application, based on the collection and processing of multiple real-time operating parameters, combined with advanced feature extraction and abnormal pattern analysis models, comprehensive monitoring and accurate fault identification of optical communication devices are achieved. Through multi-level abnormal behavior pattern recognition and real-time fault warning, the operation stability and maintenance efficiency of the equipment are effectively improved, the cost of equipment management and debugging is significantly reduced, and ultimately the user's optical communication service experience is improved.

[0053] Figure 3 Shown according to Figure 1 An example operation flow chart of step S110 in FIG.

[0054] like Figure 3 As shown, in step S310, based on a sensor group deployed in the optical communication device, a sensor observation data group of the optical communication device is collected, and the sensor group includes a temperature sensor, a current sensor, a voltage sensor and an optical power sensor.

[0055] In some embodiments, temperature and humidity sensors are deployed inside and outside the optical communication device to monitor the temperature and humidity inside the device and the ambient temperature and humidity respectively. Current sensors are installed on power lines and key circuit nodes to monitor current changes in real time. Voltage sensors are arranged at power input and output ports and key circuit nodes to monitor the voltage level when the device is running. In addition, optical power sensors placed at the optical transmitting end and the receiving end are used to monitor the changes in optical power in real time.

[0056] In addition, by establishing a high-frequency data acquisition system, it is ensured that the data of each sensor can be collected and transmitted in real time and at high speed, and the integrity and consistency of the data are ensured.

[0057] In step S320, each sensor observation data in the sensor observation data group is calibrated and combined based on the Bayesian estimation algorithm to obtain corresponding real-time operation data.

[0058] Specifically, by constructing a Bayesian estimation model and considering the error distribution and prior information of the sensor, the Bayesian estimation method is used to calibrate different types of sensor data separately to correct the systematic error and random error of the sensor and generate high-reliability real-time operation data.

[0059] In some embodiments, it is assumed that the prior distribution follows a normal distribution: ; Represents the prior distribution, which means the parameters before the observation data are available estimates; represents the prior mean, is the prior variance; Indicates is the mean, is a normal distribution with variance;

[0060] The observation data provided by each sensor can all be considered as samples from a normal distribution, and the likelihood function is ; Indicates t Observation data of sensors; is the likelihood function, which means that given the parameters Appears in the case of probability; Indicates t The mean of the observed data of the sensors, Indicates t The variance of the observation data of each sensor;

[0061] According to Bayes' theorem, update the posterior variance and posterior mean of each sensor:

[0062] , formula (1);

[0063] , formula (2);

[0064] In the formula, represents the posterior variance, Indicates the total number of sensors in the sensor group, represents the posterior mean;

[0065] The posterior mean corresponding to each sensor Combined to determine real-time operating data.

[0066] For example, assuming that the prior mean is , the prior variance is , the sensor observation data of the temperature sensor is , , and its updated variance is , the updated mean is , thus the calibrated real-time temperature sensing parameter is 16.67. Then, the real-time operation data is obtained by combining the calibrated sensing parameters.

[0067] In an embodiment of the present application, in the performance monitoring of the optical communication device, each sensor observation data in the sensor observation data group is calibrated and fused based on the Bayesian estimation algorithm to obtain the corresponding real-time operation data. Through the Bayesian estimation algorithm, the multi-sensor observation data is fused and processed, and the observation data of each sensor and its measurement error are comprehensively considered to obtain more accurate real-time operation data. Therefore, compared with the data acquisition of a single sensor, the data of each sensor is calibrated through Bayesian estimation, the measurement error and deviation between different sensors are eliminated, and the accuracy of the fused data is significantly improved, ensuring the consistency and reliability of the data, effectively reducing the monitoring data deviation caused by sensor error, and enhancing the reliability of the operation status evaluation of the optical communication device.

[0068] Figure 4 A schematic diagram of the structural connection of an example of an abnormal pattern analysis model according to an embodiment of the present application is shown.

[0069] like Figure 4 As shown, the abnormal pattern analysis model 400 includes a principal component analysis module 410 , a linear discriminant analysis module 420 and an autoencoder 430 .

[0070] More specifically, the principal component analysis module 410 is used to perform dimensionality reduction processing on the real-time operation data and extract main features to reduce the data dimension.

[0071] In some implementations, the covariance matrix corresponding to the historical running data set is calculated:

[0072] , formula (3);

[0073] In the formula, represents the covariance matrix, Indicates Historical operation data, Represents the average value of historical running data. Indicates the total number of historical operation data contained in the historical operation data set; Representation vector The transpose of .

[0074] Performing eigenvalue decomposition on the covariance matrix to obtain a plurality of first eigenvalues ​​and corresponding first eigenvectors; wherein the first eigenvalue represents the importance of the principal component, and the first eigenvector represents the direction of the principal component;

[0075] Select the one with the largest first eigenvalue The eigenvectors are used as principal components to construct the dimension reduction matrix:

[0076] , formula (4);

[0077] In the formula, represents the dimension reduction matrix, Represents the first u vector features;

[0078] The real-time operation data is subjected to dimensionality reduction processing using a dimensionality reduction matrix to obtain dimensionality-reduced feature data:

[0079] , formula (5);

[0080] In the formula, is the feature data after dimension reduction, For real-time operation data.

[0081] In this embodiment, PCA is used to perform dimensionality reduction processing on real-time operation data, retain the most representative features in the data, remove redundant and noise information, extract main features, reduce data dimensions, reduce the amount of data for subsequent processing, and improve the data representation capability.

[0082] On the other hand, the linear discriminant analysis module 420 is used to extract discriminant features from the feature data after dimension reduction to improve the separability of the features.

[0083] More specifically, the intra-class scatter matrix and the inter-class scatter matrix are calculated based on the data subsets of each category in the historical operation data set. Here, the inter-class scatter matrix is ​​used to represent the degree of dispersion between different categories, and the intra-class scatter matrix is ​​used to represent the degree of dispersion of samples within the same category. More specifically, the intra-class scatter matrix and the inter-class scatter matrix are calculated by combining the historical operation data subsets corresponding to the equipment operation status (normal or faulty) with different labels. The calculation formulas of the inter-class scatter matrix and the intra-class scatter matrix are as follows:

[0084] , formula (6);

[0085] , formula (7);

[0086] In the formula, represents the within-class scatter matrix, Indicates A subset of class historical operation data, x express Single historical running data in, For the The mean of class data, is the total number of categories corresponding to the historical running data set, Representation vector The transpose of represents the between-class scatter matrix, For the The total number of historical operation data contained in the historical operation data subset of the class, is the mean value of all data corresponding to the historical running data set, Representation vector The transpose of .

[0087] Perform eigenvalue decomposition on the product of the inverse matrix of the intra-class scatter matrix and the inter-class scatter matrix to obtain the second eigenvalue and the second eigenvector; wherein the second eigenvalue represents the importance of the discriminant feature, and the second eigenvector represents the direction of the discriminant feature;

[0088] Select the one with the largest second eigenvalue The second eigenvector is used as the discriminant feature to construct the discriminant feature matrix:

[0089] , formula (8);

[0090] In the formula, represents the discriminant feature matrix, Indicates feature vectors.

[0091] Use the discriminant feature matrix to extract discriminant features from the dimension-reduced feature data to obtain discriminant feature data:

[0092] , formula (9);

[0093] In the formula, is the discriminant feature data.

[0094] In this embodiment, based on LDA, the intra-class scatter matrix and the inter-class scatter matrix are calculated to extract features that can effectively distinguish different categories, so as to maximize the ratio of the inter-class scatter to the intra-class scatter, thereby enhancing the distinguishing ability of the features. Thus, the model's ability to distinguish different operating states or fault types is improved, and the accuracy of fault detection and classification is enhanced.

[0095] On the other hand, the autoencoder 430 is used to perform nonlinear feature learning on the discriminant feature data to extract deep-level equipment operation features.

[0096] As a further optimization of this embodiment, the autoencoder 430 includes an encoder module 431 and a decoder 433. The encoder module 431 includes a plurality of encoders, each of which has a unique corresponding scale to capture feature information of different scales through a parallel structure. A self-attention mechanism is introduced in the encoder of each scale to automatically adjust the importance weight of the feature by calculating the correlation between different features.

[0097] In this embodiment, the discriminant feature data is used as input to train the multi-scale autoencoder, so that the low-dimensional features output by the encoder can obtain effective feature representation data through nonlinear feature extraction. The multi-scale feature extraction mechanism and self-attention mechanism are introduced in the autoencoder module. Through the multi-level encoder structure, the parallel structure is used to capture feature information of different scales, which can improve the expressiveness of the features and better reflect the operating status of the equipment. In addition, the self-attention mechanism is introduced in the encoder of each scale. By calculating the correlation between different features and automatically adjusting the importance weight of the features, the model can adaptively select the most representative features under different operating conditions, improve the adaptability of the model under different operating conditions and fault types, and make feature extraction more flexible and efficient.

[0098] Figure 5 An operational flowchart of an example of identifying a first abnormal behavior pattern by an association rule mining module according to an embodiment of the present application is shown.

[0099] like Figure 5 As shown, in step S510, association rule mining is performed at the device level to identify device-level association rules between the overall device operation characteristics of the optical communication device and the first calibrated abnormal behavior pattern.

[0100] In some implementations, common abnormal device behavior patterns, such as sudden drop in optical power, abnormal temperature rise, etc., are identified and recorded through historical data and expert knowledge, so as to achieve learning and foundation for abnormal device behavior patterns. In addition, by analyzing historical failure data of different components inside the device, abnormal behavior patterns of components are identified and recorded, such as excessive laser temperature, decreased sensitivity of optical receivers, etc., so as to achieve learning and foundation for abnormal component behavior patterns.

[0101] Exemplarily, assuming that the operating characteristics of the optical communication device include temperature, current, voltage and optical power, the association between these characteristics and the overall operating state of the device can be mined to identify the association between the overall operating characteristics of the optical communication device and abnormal behavior patterns.

[0102] By mining association rules at the device level, the association between the overall operating characteristics of the optical communication device and the abnormal behavior pattern can be identified, ensuring that the operating status of the device is grasped as a whole and identifying the association rules with the first calibrated abnormal behavior pattern.

[0103] In step S520, association rule mining is performed at the component level to identify component-level association rules between component operation characteristics of each component inside the optical communication device and the second calibrated abnormal behavior pattern.

[0104] Exemplarily, assuming that the optical communication device includes a laser, an optical receiver, and an optical amplifier, association rule mining can be performed at the component level to identify the association between the operating characteristics and abnormal behavior patterns of different components within the optical communication device.

[0105] Association rule mining at the component level can identify the association rules between the component operation characteristics of each component inside the optical communication device and the second standard abnormal behavior pattern, and can more carefully analyze the operation status of each component inside the device and identify the association rules with the second standard abnormal behavior pattern.

[0106] In step S530, it is identified whether a first abnormal behavior pattern exists according to the device-level association rule and the component-level association rule.

[0107] Here, by integrating the association rules at the device level and the component level and based on multi-level association rule mining operations, it is possible to ensure that the operating characteristics of the optical communication device are fully covered from the overall to the local, from the device as a whole to the component level, and to fully capture the abnormal behavior patterns of the equipment, thus ensuring the comprehensiveness and accuracy of fault identification.

[0108] Figure 6 An operational flow chart of an example of identifying a second abnormal behavior pattern by a cluster analysis module according to an embodiment of the present application is shown.

[0109] like Figure 6 As shown, in step S610, each data point in the equipment operation characteristic data is dynamically clustered with the initial cluster group based on the adaptive K-means algorithm. The initial cluster group is set to have a preset number of initial clusters, and each initial cluster has a corresponding center point.

[0110] In some embodiments, before cluster analysis is performed, the initial number of clusters and cluster centers need to be determined. Further, multiple possible cluster centers are pre-set through historical data analysis and expert knowledge. In addition, the adaptive K-means algorithm will automatically adjust the number of clusters based on the initial cluster centers and the distribution of data to find the optimal number of clusters. In this embodiment, the number of clusters is dynamically adjusted by the silhouette coefficient to find the optimal number of clusters.

[0111] More specifically, the calculation formula of the silhouette coefficient is:

[0112] , formula (10);

[0113] In the formula, Data points in the device operation characteristic data The average distance to other points in the same cluster, For data points The average distance to the nearest cluster, The total number of data points in the equipment operation characteristic data; G Represents the silhouette coefficient.

[0114] In step S620, for the clustering result corresponding to the maximum silhouette coefficient, the local outlier factor value of each data point is analyzed.

[0115] More specifically, based on the dynamic clustering results, the local outlier factor (LOF) method is used for local anomaly detection. The local density of each data point and the density of its adjacent points are calculated, and the local density differences are compared to identify abnormal data points in the cluster.

[0116] More specifically, the calculation formula for the LOF value is:

[0117] , formula (11);

[0118] In the formula, Represents a data point in the equipment operation characteristic data The local anomaly factor value of For data points of Neighborhood, For data points The local reachable density of For data points The local reachable density of .

[0119] In step S630, the abnormal data point distribution is determined based on the calculated local abnormal factor value and each data point, and whether a second abnormal behavior pattern exists is identified based on the abnormal data point distribution, where the abnormal data point distribution includes the number, location and aggregation of the abnormal data points.

[0120] In some embodiments, by marking some data points with higher or prominent corresponding LOF values ​​as abnormal data points, for example, the LOF value exceeds a preset threshold, the abnormal behavior pattern is analyzed according to the distribution of abnormal data points, for example, if the abnormal data points are concentrated in certain areas or under certain specific characteristics (such as high temperature, high current, etc.), it indicates that there is an abnormal behavior pattern.

[0121] In this embodiment, an adaptive K-means algorithm is used for dynamic clustering. By calculating the silhouette coefficient, the number of clusters can be dynamically adjusted until the silhouette coefficient of the clustering result is maximized, thereby ensuring the optimality of the clustering result, so that the compactness of each data point within its cluster and the separation from other clusters are optimal. In addition, based on the optimal clustering result, by calculating the LOF value of each data point, local abnormal points can be accurately identified, and subtle abnormalities in the operation of the equipment can be captured, especially those local abnormalities that are difficult to detect as a whole. It can accurately capture abnormalities in the operation of the equipment, further improving the accuracy of fault warning.

[0122] Figure 7 An operational flow chart of an example of determining abnormal behavior pattern characteristics according to an embodiment of the present application is shown.

[0123] like Figure 7 As shown, in step S710, when the first abnormal behavior pattern and / or the second abnormal behavior pattern is detected, the first abnormal behavior feature is determined according to the device-level association rule and the component-level association rule.

[0124] It should be noted that association rule mining itself has strong logic and interpretability, and can clearly describe the relationship between equipment operation characteristics and abnormal behavior patterns. Through this transparent mining method, when an anomaly is identified, the features mined in this way can effectively provide effective support for the reasoning and prediction of equipment failure types.

[0125] In step S720, an abnormal behavior feature set corresponding to the abnormal data point distribution is extracted, and a second abnormal behavior feature is determined according to the abnormal behavior feature set.

[0126] It should be noted that the cluster analysis module comprehensively analyzes the characteristics of each abnormal data point to form an abnormal behavior feature set, conducts detailed analysis on the data points marked as abnormal, and extracts the common features of these data points. For example, abnormal data points of high temperature and high current may together constitute an abnormal behavior feature. These common features are counted and summarized to form an abnormal behavior feature set. This can grasp the abnormal patterns in the operation of the equipment as a whole and form a comprehensive description of the abnormal behavior features.

[0127] In step S730, an abnormal behavior pattern feature is determined according to the first abnormal behavior feature and the second abnormal behavior feature.

[0128] In some embodiments, the first abnormal behavior feature and the second abnormal behavior feature are combined to determine the corresponding abnormal behavior pattern feature, and input it into the fault identification model. Through multi-level association rule mining and cluster analysis, the accuracy and comprehensiveness of the input features of the fault identification model are ensured, so that the corresponding target fault type can be effectively predicted.

[0129] In some examples of the embodiments of the present application, the fault identification model adopts an enhanced deep belief network (DBN). Figure 8 A schematic diagram of the structural connection of an example of a fault identification model according to an embodiment of the present application is shown.

[0130] like Figure 8 As shown, the fault identification model 800 includes a multimodal feature fusion layer 810 , a feature extraction layer 820 , an adaptive learning layer 830 and a classification layer 840 .

[0131] The multimodal feature fusion layer 810 is used to fuse real-time operation data, equipment operation features and abnormal behavior pattern features to obtain corresponding multimodal fusion features.

[0132] In the operation data of optical communication devices, there may be complex relationships between different parameters. Through multimodal feature fusion, features from different sources are fused to improve the expressiveness of features, capture the correlation between different parameters, and improve the fault identification ability of the model.

[0133] More specifically, it can be determined by:

[0134] , formula (12);

[0135] In the formula, represents multimodal fusion features, They represent real-time operation data, equipment operation characteristics, and abnormal behavior pattern characteristics respectively; Represents a feature concatenation operation.

[0136] Through the multimodal feature fusion layer, the fault prediction model can effectively combine data from multiple sensors, including internal temperature of the equipment, ambient temperature and humidity, real-time current, real-time voltage and optical power, enabling the model to extract more useful information in the high-dimensional data space, improving the comprehensive understanding of the equipment's operating status and ensuring that the model can maintain efficient performance in different operating environments and conditions.

[0137] The feature extraction layer 820 is used to process the multimodal fusion features through a multi-layer restricted Boltzmann machine (RBM) to obtain a deep feature matrix.

[0138] Here, DBN can extract deep features from the data layer by layer by stacking multiple RBM layers for multi-layer feature extraction. In the fault identification of optical communication devices, multi-layer feature extraction can help the model better understand the complex patterns in the data and improve the accuracy of fault identification. Using the deep belief network (DBN) as the backbone structure of the optical communication device fault identification model can fully utilize the advantages of DBN in processing high-dimensional complex data, powerful feature learning capabilities, layer-by-layer training to prevent gradient disappearance, and combining pre-training with fine-tuning.

[0139] More specifically, the process of performing data processing in the RBM layer mainly includes the process of forward propagation from the fused features to the hidden layer, and the process of reconstruction from the hidden layer to the fused features.

[0140] The process of forward propagation from fusion features to hidden layers is as follows:

[0141] , formula (13);

[0142] Thus, the hidden layer nodes are calculated The probability of being activated is calculated by Perform weighted summation and then pass the activation function Obtained by conversion.

[0143] The process of reconstruction from hidden layer to fusion feature is as follows:

[0144] , formula (14);

[0145] , formula (15);

[0146] In the formula, Indicates that given the fusion feature In the case of j Hidden layer nodes Probability of activation; Indicates the first Features, Indicates the number of fused features; Represents fusion features and hidden layer nodes The weight between Represents the hidden layer nodes Bias of ; sigmoid represents the sigmoid activation function; Indicates that at a given hidden layer node In the case of The probability of being 1; Represents input features Bias of Represents the deep features output by the feature extraction layer; Represents the total number of hidden layer nodes; Represents the set of hidden layer nodes, which contains all hidden layer nodes The activation value of .

[0147] In this embodiment, through the hidden layer nodes The activation value is used to reconstruct the input features , which can help the model calibrate and adjust weights, thereby improving the quality of feature representation.

[0148] Through multi-layer restricted Boltzmann machine (RBM) layers, layer-by-layer unsupervised learning is used to extract deep information from fusion features, which can capture complex nonlinear relationships in input data. By learning and adjusting weights, the RBM layer improves the quality of feature representation, enabling the model to identify potential failure modes, so that when processing complex equipment operation data, it can more accurately capture abnormal signals, thereby improving the accuracy of fault prediction.

[0149] The adaptive learning layer 830 is used to perform weighted fusion of the multimodal fusion features and the deep feature matrix to obtain an adaptive feature matrix.

[0150] Specifically, through the adaptive learning mechanism, the sensitivity of the model to different features is enhanced, and the effectiveness of feature extraction is improved.

[0151] Here, the adaptive learning layer 830 performs weighted fusion of the output features of the multimodal feature fusion layer and the output features of the RBM layer by dynamically adjusting the adaptive coefficients. The adaptive learning mechanism enables the model to dynamically adjust the weights of features according to the importance of the features, thereby improving the effectiveness of feature extraction and the robustness of the model. Through adaptive learning, the model can automatically adapt to the changes in the importance of different features, thereby maintaining efficient fault prediction performance under different operating environments and equipment conditions.

[0152] , formula (16);

[0153] , formula (17);

[0154] , formula (18);

[0155] In the formula, The adaptive feature matrix representing the output of the adaptive learning layer; and They represent the first The deep features and Deep features, B Represents the total number of deep features in the deep feature matrix; represents the exponential function; and Respectively for The adaptive coefficient.

[0156] In this embodiment, by introducing an adaptive learning layer and a multi-layer RBM, the fault prediction model can maintain high robustness and generalization ability when facing data from different devices and different operating environments, so that the model can adapt to various complex operating conditions, provide stable fault prediction performance, and improve the intelligence level of equipment management.

[0157] The classification layer 840 is used to process the adaptive feature matrix to output the corresponding target fault type:

[0158] , formula (19);

[0159] , formula (20);

[0160] In the formula, represents the intermediate feature matrix output by the fully connected layer, and Respectively represent the first Features and Features, H Represents the total number of features in the intermediate feature matrix; and Represent the weight matrix and bias term of the fully connected layer respectively; When represents the input feature, the predicted fault type is probability.

[0161] In this embodiment, the classification layer maps the features extracted by the adaptive learning layer to a specific fault type space through a fully connected layer and a Softmax layer, which can accurately classify the fault types of the optical communication device, provide clear fault prediction results, guide maintenance personnel to handle faults in a timely manner, and improve the maintenance efficiency of the optical communication device.

[0162] Through the above-mentioned model structure design of the embodiment of the present application, the fault prediction model can detect potential failure modes in time before the failure occurs, provide early warning, and enable maintenance personnel to take preventive measures before the failure occurs. It can effectively reduce communication interruptions caused by equipment failures, improve users' service experience, and reduce downtime and maintenance costs.

[0163] In some examples of the embodiments of the present application, in the fault identification scenario of an optical communication device, the loss function of the fault identification model may adopt a comprehensive loss function that integrates basic classification loss, imbalanced data adjustment, and misjudgment cost. More specifically, the loss function of the fault identification model is It is expressed by the following formula:

[0164] , formula (21);

[0165] , formula (22);

[0166] , formula (22);

[0167] , formula (23);

[0168] In the formula, represents the cross entropy loss, represents the imbalanced data loss term, and It represents the misjudgment cost loss item; , and Respectively represent the loss adjustment weights corresponding to each loss item; is the total number of samples in the data sample set, is the total number of fault categories; Indicates The true label of the sample, if samples belong to the category ,but ,otherwise ; The model predicts samples belong to the category probability; η Represents the adjustment parameter, which is used to control the degree of attention the model pays to this; Indicates that the samples are misclassified as The value of time.

[0169] In the embodiment of the present application, by adopting the cross-entropy loss as the basic classification loss, the classification accuracy of the model in the multi-classification task is ensured, so that the model can accurately identify different types of faults, reduce false positives and false negatives, and ensure the normal operation of the optical communication device. In the unbalanced data loss term, by introducing the adjustment factor (Focal Loss), the model can enhance the focus on difficult-to-classify samples, especially for unbalanced data. This enables the model to maintain efficient recognition capabilities when facing a small number of class faults with a small amount of data. In the misjudgment cost loss term, the misjudgment cost (MisclassificationCost) is introduced to weight the misjudgment of different fault types, especially setting a higher cost for critical faults, reducing the negative impact of misjudgment on system stability and security, and improving the reliability of fault identification. Furthermore, by adjusting the weights of each loss, the basic classification loss, unbalanced data adjustment and misjudgment cost are used to comprehensively determine the loss function, so that the model can optimize multiple objectives at the same time during the training process, improve the overall performance, and ensure the efficiency and robustness of the model in practical applications.

[0170] It should be noted that, for the aforementioned method embodiments, for the sake of simplicity of description, they are all expressed as a series of actions combined, but those skilled in the art should be aware that the present application is not limited by the described order of actions, because according to the present application, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the present application. In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, please refer to the relevant description of other embodiments.

[0171] Fig. 9 A structural block diagram of an example of an optical communication device performance monitoring system based on big data analysis according to an embodiment of the present application is shown.

[0172] like Fig. 9 As shown, the optical communication device performance monitoring system 900 based on big data analysis includes an operation data acquisition unit 910, an operation feature extraction unit 920, a feature pattern analysis unit 930, a fault type analysis unit 940 and a performance indicator monitoring unit 950.

[0173] The operation data acquisition unit 910 is used to acquire real-time operation data of the optical communication device; the real-time operation data includes multiple types of operation parameters: internal temperature of the device, ambient temperature and humidity, real-time current, real-time voltage and optical power.

[0174] The operation feature extraction unit 920 is used to extract the equipment operation features corresponding to the real-time operation data based on the feature extraction model; the feature extraction model includes a principal component analysis module, a linear discriminant analysis module and an autoencoder, the principal component analysis module is used to perform dimensionality reduction processing on the real-time operation data, and then perform discriminant feature extraction through the linear discriminant analysis module, and then perform nonlinear feature learning through the autoencoder to obtain the corresponding equipment operation features.

[0175] The characteristic pattern analysis unit 930 is used to extract the abnormal behavior pattern characteristics corresponding to the equipment operation characteristics based on the abnormal pattern analysis model, specifically including:

[0176] The abnormal pattern analysis model includes an association rule mining module and a cluster analysis module; the association rule mining module is used to perform association rule mining on the device operation feature data to identify whether there is a first abnormal behavior pattern; the cluster analysis module is used to perform cluster analysis on the device operation feature data to identify whether there is a second abnormal behavior pattern.

[0177] When the first abnormal behavior pattern and / or the second abnormal behavior pattern is detected, the abnormal behavior pattern features are determined according to the first abnormal behavior features obtained by the association rule mining and the second abnormal behavior features obtained by the cluster analysis.

[0178] The fault type analysis unit 940 is used to input the real-time operation data, the equipment operation characteristics and the abnormal behavior pattern characteristics into the fault identification model, so that the fault identification model outputs the corresponding target fault type; the target fault type includes any one of the following: laser aging fault, laser temperature drift fault, laser drive circuit fault, optical receiver sensitivity fault and optical amplifier gain abnormality.

[0179] The performance indicator monitoring unit 950 is used to output the equipment performance indicators of the optical communication device according to the real-time operation data when the first abnormal behavior pattern or the second abnormal behavior pattern is not detected; the equipment performance indicators include equipment operation stability, equipment energy efficiency ratio, signal quality ratio and power conversion efficiency.

[0180] In some embodiments, an embodiment of the present application provides a non-volatile computer-readable storage medium, in which one or more programs including execution instructions are stored, and the execution instructions can be read and executed by an electronic device (including but not limited to a computer, a server, or a network device, etc.) to execute the steps of any of the above-mentioned methods for monitoring the performance of an optical communication device based on big data analysis in the present application.

[0181] In some embodiments, the embodiments of the present application also provide a computer program product, which includes a computer program stored on a non-volatile computer-readable storage medium, and the computer program includes program instructions. When the program instructions are executed by a computer, the computer performs any step of the above-mentioned method for monitoring the performance of an optical communication device based on big data analysis.

[0182] In some embodiments, the embodiments of the present application also provide an electronic device, comprising: at least one processor, and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the steps of the optical communication device performance monitoring method based on big data analysis.

[0183] Fig.10 is a schematic diagram of the hardware structure of an electronic device for performing a method for monitoring the performance of an optical communication device based on big data analysis provided by another embodiment of the present application, such as Fig.10 As shown, the device includes:

[0184] One or more processors 1010 and memory 1020, Fig.10 A processor 1010 is taken as an example.

[0185] The apparatus for executing the optical communication device performance monitoring method based on big data analysis may further include: an input device 1030 and an output device 1040 .

[0186] The processor 1010, the memory 1020, the input device 1030 and the output device 1040 may be connected via a bus or other means. Fig.10 The example of connecting through bus is taken in the following.

[0187] The memory 1020, as a non-volatile computer-readable storage medium, can be used to store non-volatile software programs, non-volatile computer executable programs and modules, such as program instructions / modules corresponding to the optical communication device performance monitoring method based on big data analysis in the embodiment of the present application. The processor 1010 executes various functional applications and data processing of the server by running the non-volatile software programs, instructions and modules stored in the memory 1020, that is, the optical communication device performance monitoring method based on big data analysis in the above method embodiment is implemented.

[0188] The memory 1020 may include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function; the data storage area may store data created according to the use of the electronic device, etc. In addition, the memory 1020 may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other non-volatile solid-state storage device. In some embodiments, the memory 1020 may optionally include a memory remotely arranged relative to the processor 1010, and these remote memories may be connected to the electronic device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0189] The input device 1030 may receive input digital or character information and generate signals related to user settings and function control of the electronic device. The output device 1040 may include a display device such as a display screen.

[0190] The one or more modules are stored in the memory 1020, and when executed by the one or more processors 1010, perform the optical communication device performance monitoring method based on big data analysis in any of the above method embodiments.

[0191] The above-mentioned product can execute the method provided in the embodiment of the present application, and has the functional modules and beneficial effects corresponding to the execution method. For technical details not fully described in this embodiment, please refer to the method provided in the embodiment of the present application.

[0192] The electronic devices of the embodiments of the present application exist in various forms, including but not limited to:

[0193] (1) Mobile communication equipment: This type of equipment is characterized by having mobile communication functions and its main purpose is to provide voice and data communications. This type of terminal includes: smart phones, multimedia phones, functional phones, and low-end phones.

[0194] (2) Ultra-mobile personal computer devices: These devices belong to the category of personal computers, have computing and processing functions, and generally also have mobile Internet access features. These terminals include: PDA, MID and UMPC devices, etc.

[0195] (3) Portable entertainment devices: These devices can display and play multimedia content. They include audio and video players, handheld game consoles, e-books, smart toys, and portable car navigation devices.

[0196] (4) Other onboard electronic devices with data interaction functions, such as on-board devices installed in vehicles.

[0197] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0198] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a general hardware platform, and of course, by hardware. Based on this understanding, the above technical solution, in essence, or the part that contributes to the relevant technology, can be embodied in the form of a software product, which can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiment.

[0199] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for monitoring the performance of an optical communication device based on big data analysis, comprising: Obtaining real-time operation data of optical communication devices; The real-time operation data includes multiple types of operation parameters: internal temperature of the device, ambient temperature and humidity, real-time current, real-time voltage and optical power; Extracting the equipment operation features corresponding to the real-time operation data based on the feature extraction model; the feature extraction model comprises a principal component analysis module, a linear discriminant analysis module and an autoencoder, the principal component analysis module is used to perform dimensionality reduction processing on the real-time operation data, and then perform discriminant feature extraction through the linear discriminant analysis module, and then perform nonlinear feature learning through the autoencoder, so as to obtain the corresponding equipment operation feature data; Extracting abnormal behavior pattern features corresponding to the equipment operation features based on the abnormal pattern analysis model specifically includes: The abnormal pattern analysis model includes an association rule mining module and a cluster analysis module; the association rule mining module is used to perform association rule mining on the device operation feature data to identify whether there is a first abnormal behavior pattern; the cluster analysis module is used to perform cluster analysis on the device operation feature data to identify whether there is a second abnormal behavior pattern. In the case where the first abnormal behavior pattern and / or the second abnormal behavior pattern is detected, determining abnormal behavior pattern features according to the first abnormal behavior features obtained by the association rule mining and the second abnormal behavior features obtained by the cluster analysis; The real-time operation data, the equipment operation characteristics and the abnormal behavior pattern characteristics are input into a fault identification model, so that the fault identification model outputs a corresponding target fault type; the target fault type includes any one of the following: laser aging fault, laser temperature drift fault, laser drive circuit fault, optical receiver sensitivity fault and optical amplifier gain abnormality; When the first abnormal behavior mode or the second abnormal behavior mode is not detected, outputting device performance indicators of the optical communication device according to the real-time operation data; the device performance indicators include device operation stability, device energy efficiency ratio, signal quality ratio and power conversion efficiency; Wherein, the obtaining of real-time operation data of the optical communication device includes: Based on a sensor group deployed in the optical communication device, a sensor observation data group of the optical communication device is collected; the sensor group includes a temperature sensor, a current sensor, a voltage sensor and an optical power sensor; Based on the Bayesian estimation algorithm, each sensor observation data in the sensor observation data group is calibrated and combined to obtain corresponding real-time operation data, specifically including: Assume that the prior distribution follows a normal distribution: ; Represents the prior distribution, which means the parameters before the observation data are available estimates; represents the prior mean, is the prior variance; Indicates is the mean, is a normal distribution with variance; The observation data provided by each sensor are all considered to be sampled from a normal distribution, and the likelihood function is ; Indicates Observation data of sensors; is the likelihood function, which means that given the parameters Appears in the case of probability; Indicates The mean of the observed data of the sensors, Indicates The variance of the observation data of each sensor; According to Bayes' theorem, update the posterior variance and posterior mean of each sensor: In the formula, represents the posterior variance, Indicates the total number of sensors in the sensor group, represents the posterior mean; The posterior mean corresponding to each sensor Combined to determine real-time operating data.

2. The method according to claim 1, wherein: The principal component analysis module is used to perform dimensionality reduction processing on real-time operation data and extract main features to reduce data dimensions: Calculate the covariance matrix corresponding to the historical running data set: In the formula, represents the covariance matrix, Indicates Historical operation data, Represents the average value of historical running data. Indicates the total number of historical operation data contained in the historical operation data set; Representation vector The transpose of Performing eigenvalue decomposition on the covariance matrix to obtain a plurality of first eigenvalues ​​and corresponding first eigenvectors; wherein the first eigenvalue represents the importance of the principal component, and the first eigenvector represents the direction of the principal component; Select the first eigenvalue after sorting from largest to smallest The eigenvectors are used as principal components to construct the dimension reduction matrix: In the formula, represents the dimension reduction matrix, Represents the first vector features; The real-time operation data is subjected to dimensionality reduction processing using a dimensionality reduction matrix to obtain dimensionality-reduced feature data: In the formula, is the feature data after dimension reduction, To run data in real time; The linear discriminant analysis module is used to extract discriminant features from the feature data after dimension reduction to improve the separability of the features: The intra-class scatter matrix and the inter-class scatter matrix are calculated based on the data subsets of each category in the historical running data set: In the formula, represents the within-class scatter matrix, Indicates A subset of class historical operation data, express Single historical running data in, For the The mean of class data, is the total number of categories corresponding to the historical running data set, Representation vector The transpose of represents the between-class scatter matrix, For the The total number of historical operation data contained in the historical operation data subset of the class, is the mean value of all data corresponding to the historical running data set, Representation vector The transpose of Perform eigenvalue decomposition on the product of the inverse matrix of the intra-class scatter matrix and the inter-class scatter matrix to obtain the second eigenvalue and the second eigenvector; wherein the second eigenvalue represents the importance of the discriminant feature, and the second eigenvector represents the direction of the discriminant feature; Select the first eigenvalue corresponding to the second eigenvalue sorted from largest to smallest The second eigenvector is used as the discriminant feature to construct the discriminant feature matrix: In the formula, represents the discriminant feature matrix, Indicates feature vectors; Use the discriminant feature matrix to extract discriminant features from the dimension-reduced feature data to obtain discriminant feature data: In the formula, is the discriminant feature data.

3. The method according to claim 2, wherein: The autoencoder is used to perform nonlinear feature learning on the discriminant feature data to extract deep-level equipment operation features; The autoencoder includes an encoder module and a decoder, the encoder module includes multiple encoders, each of which has a unique corresponding scale to capture feature information of different scales through a parallel structure; a self-attention mechanism is introduced in the encoder of each scale to automatically adjust the importance weight of the feature by calculating the correlation between different features.

4. The method according to any one of claims 1 to 3, wherein: The association rule mining module is used to perform multi-level rule mining operations, specifically including: Performing association rule mining at the device level to identify device-level association rules between the overall device operation characteristics of the optical communication device and the first calibrated abnormal behavior pattern; Performing association rule mining at the component level to identify component-level association rules between component operation characteristics of each component within the optical communication device and the second calibrated abnormal behavior pattern; According to the device-level association rule and the component-level association rule, it is identified whether a first abnormal behavior pattern exists.

5. The method according to claim 4, wherein: The cluster analysis module is used to identify whether there is a second abnormal behavior pattern through the following operations: Dynamically clustering each data point in the equipment operation characteristic data with an initial cluster group based on an adaptive K-means algorithm; the initial cluster group is set to have a preset number of initial clusters, and each initial cluster has a corresponding center point; The calculation formula of silhouette coefficient is: In the formula, Data points in the device operation characteristic data The average distance to other points in the same cluster, For data points The average distance to the nearest cluster, The total number of data points in the equipment operation characteristic data; represents the silhouette coefficient; For the clustering result with the largest silhouette coefficient, analyze the local anomaly factor value of each data point: In the formula, Represents a data point in the equipment operation characteristic data The local anomaly factor value of For data points of Neighborhood, For data points The local reachable density of For data points The local reachable density of The abnormal data point distribution is determined according to the calculated local abnormal factor value and each data point, and whether a second abnormal behavior pattern exists is identified according to the abnormal data point distribution; the abnormal data point distribution includes the number, position and aggregation of the abnormal data points.

6. The method according to claim 5, wherein: In the case where the first abnormal behavior pattern and / or the second abnormal behavior pattern is detected, determining the abnormal behavior pattern feature according to the first abnormal behavior feature obtained by the association rule mining and the second abnormal behavior feature obtained by the cluster analysis includes: In case the first abnormal behavior pattern and / or the second abnormal behavior pattern are detected, determining a first abnormal behavior feature according to the device-level association rule and the component-level association rule; and Extracting an abnormal behavior feature set corresponding to the abnormal data point distribution, and determining a second abnormal behavior feature according to the abnormal behavior feature set; An abnormal behavior pattern feature is determined according to the first abnormal behavior feature and the second abnormal behavior feature.

7. The method according to claim 1, wherein: The fault identification model adopts an enhanced deep belief network, which includes a multimodal feature fusion layer, a feature extraction layer, an adaptive learning layer and a classification layer; The multimodal feature fusion layer is used to fuse the real-time operation data, the equipment operation features and the abnormal behavior pattern features to obtain corresponding multimodal fusion features: In the formula, represents multimodal fusion features, They represent real-time operation data, equipment operation characteristics, and abnormal behavior pattern characteristics respectively; Represents feature concatenation operation; The feature extraction layer is used to process the multimodal fusion features through a multi-layer restricted Boltzmann machine RBM to obtain a deep feature matrix: In the formula, Indicates that given the fusion feature In the case of Hidden layer nodes Probability of activation; Indicates the first Features, Indicates the number of fused features; Represents fusion features and hidden layer nodes The weight between Represents the hidden layer nodes Bias of Represents the sigmoid activation function; Indicates that at a given hidden layer node In the case of The probability of being 1; Represents input features Bias of Represents the deep features output by the feature extraction layer; Represents the total number of hidden layer nodes; Represents the set of hidden layer nodes, which contains all hidden layer nodes The activation value of The adaptive learning layer is used to perform weighted fusion of the multimodal fusion features and the deep feature matrix to obtain an adaptive feature matrix: In the formula, The adaptive feature matrix representing the output of the adaptive learning layer; and They represent the first The deep features and A deep feature, Represents the total number of deep features in the deep feature matrix; represents the exponential function; and Respectively for The adaptive coefficient of The classification layer is used to process the adaptive feature matrix to output the corresponding target fault type: In the formula, represents the intermediate feature matrix output by the fully connected layer, and Respectively represent the first Features and Features, Represents the total number of features in the intermediate feature matrix; and Represent the weight matrix and bias term of the fully connected layer respectively; Represents input features When , the predicted fault type is probability.

8. The method according to claim 7, wherein: The loss function of the fault identification model is for: In the formula, represents the cross entropy loss, represents the imbalanced data loss term, and It represents the misjudgment cost loss item; , and Respectively represent the loss adjustment weights corresponding to each loss item; is the total number of samples in the data sample set, is the total number of fault categories; Indicates The true label of the sample, if samples belong to the category ,but ,otherwise ; The model predicts samples belong to the category probability; Represents the adjustment parameter, which is used to control the degree of attention the model pays to this; Indicates that the samples are misclassified as The value of time.

9. An optical communication device performance monitoring system based on big data analysis, comprising: An operation data acquisition unit, used to acquire real-time operation data of the optical communication device; The real-time operation data includes multiple types of operation parameters: internal temperature of the device, ambient temperature and humidity, real-time current, real-time voltage and optical power; An operation feature extraction unit is used to extract the equipment operation features corresponding to the real-time operation data based on a feature extraction model; the feature extraction model includes a principal component analysis module, a linear discriminant analysis module and an autoencoder, the principal component analysis module is used to perform dimensionality reduction processing on the real-time operation data, and then perform discriminant feature extraction through the linear discriminant analysis module, and then perform nonlinear feature learning through the autoencoder to obtain corresponding equipment operation feature data; The characteristic pattern analysis unit is used to extract the abnormal behavior pattern characteristics corresponding to the equipment operation characteristics based on the abnormal pattern analysis model, specifically including: The abnormal pattern analysis model includes an association rule mining module and a cluster analysis module; the association rule mining module is used to perform association rule mining on the device operation feature data to identify whether there is a first abnormal behavior pattern; the cluster analysis module is used to perform cluster analysis on the device operation feature data to identify whether there is a second abnormal behavior pattern. In the case where the first abnormal behavior pattern and / or the second abnormal behavior pattern is detected, determining abnormal behavior pattern features according to the first abnormal behavior features obtained by the association rule mining and the second abnormal behavior features obtained by the cluster analysis; A fault type analysis unit, used for inputting the real-time operation data, the equipment operation characteristics and the abnormal behavior pattern characteristics into a fault identification model, so that the fault identification model outputs a corresponding target fault type; the target fault type includes any one of the following: laser aging fault, laser temperature drift fault, laser drive circuit fault, optical receiver sensitivity fault and optical amplifier gain abnormality; a performance indicator monitoring unit, configured to output, when the first abnormal behavior mode or the second abnormal behavior mode is not detected, a device performance indicator of the optical communication device according to the real-time operation data; the device performance indicator comprises device operation stability, device energy efficiency ratio, signal quality ratio and power conversion efficiency; Wherein, the obtaining of real-time operation data of the optical communication device includes: Based on a sensor group deployed in the optical communication device, a sensor observation data group of the optical communication device is collected; the sensor group includes a temperature sensor, a current sensor, a voltage sensor and an optical power sensor; Based on the Bayesian estimation algorithm, each sensor observation data in the sensor observation data group is calibrated and combined to obtain corresponding real-time operation data, specifically including: Assume that the prior distribution follows a normal distribution: ; Represents the prior distribution, which means the parameters before the observation data are available estimates; represents the prior mean, is the prior variance; Indicates is the mean, is a normal distribution with variance; The observation data provided by each sensor are all considered to be sampled from a normal distribution, and the likelihood function is ; Indicates Observation data of sensors; is the likelihood function, which means that given the parameters In the case of probability; Indicates The mean of the observed data of the sensors, Indicates The variance of the observation data of each sensor; According to Bayes' theorem, update the posterior variance and posterior mean of each sensor: In the formula, represents the posterior variance, Indicates the total number of sensors in the sensor group, represents the posterior mean; The posterior mean corresponding to each sensor Combined to determine real-time operating data.

Images