Abnormal recognition method and device, computer equipment, readable storage medium and program product

Abstract

The invention relates to an exception recognition method and device, computer equipment, a computer readable storage medium and a computer program product. The method comprises the steps that training data are acquired, the training data comprise multiple pieces of sample data and sample anomaly types corresponding to the sample data, and the sample data are amplitude information and phase information of vibration signals collected at a target position at a target moment; processing the training data through the initial exception type recognition model to obtain a predicted exception type corresponding to the sample data; and training the initial anomaly type recognition model based on the predicted anomaly type and the sample anomaly type to obtain a trained target anomaly type recognition model. By adopting the method, the training cost can be reduced, the calculation energy efficiency can be improved, the complexity of model processing can be reduced, the efficiency of network operation and maintenance can be further improved, the calculation time of external damage identification can be shortened, and the analysis capability and the abnormal condition positioning capability of abnormal conditions of optical fibers and cables can be improved.

Description

Technical Field

[0001] The present application relates to the field of optical cable technology, and in particular to an abnormality identification method, device, computer equipment, computer-readable storage medium and computer program product. Background Art

[0002] The related technologies have risks of cable co-routing such as capacity expansion, cutover, label loss, and external damage to cables caused by construction and other human damage, which requires management of optical network resources and identification of abnormal situations. Generally, abnormalities are identified through artificial intelligence-based recognition algorithms, but since the artificial intelligence-based recognition algorithms require a lot of learning costs, the complexity of the recognition algorithms is high, resulting in low computational efficiency of the algorithms. Summary of the invention

[0003] Based on this, it is necessary to provide an anomaly identification method, device, computer equipment, computer-readable storage medium and computer program product that can reduce the complexity of the identification algorithm and improve computing efficiency in response to the above technical problems.

[0004] In a first aspect, the present application provides an abnormality identification method, comprising:

[0005] Acquire training data, wherein the training data includes a plurality of sample data and a sample abnormality type corresponding to each of the sample data, wherein the sample data is amplitude information and phase information of a vibration signal collected at a target position at a target time;

[0006] Processing the training data through an initial anomaly type recognition model to obtain a predicted anomaly type corresponding to the sample data;

[0007] Based on the predicted abnormality type and the sample abnormality type, the initial abnormality type recognition model is trained to obtain a trained target abnormality type recognition model.

[0008] In one embodiment, the initial abnormality type recognition model is trained based on the predicted abnormality type and the sample abnormality type to obtain a trained target abnormality type recognition model, including:

[0009] Obtaining a loss value based on the loss function, the predicted anomaly type, and the sample anomaly type;

[0010] When the preset training completion conditions are not met, the weight parameters of the initial abnormality type recognition model are updated based on the loss value to obtain an updated abnormality type recognition model, and based on the updated abnormality type recognition model, the step of obtaining training data is re-executed until the preset training completion conditions are met to obtain a trained target abnormality type recognition model.

[0011] In one embodiment, the updating of the weight parameters corresponding to each layer of the initial abnormality type recognition model based on the loss value to obtain an updated abnormality type recognition model includes:

[0012] Based on the loss value, the gradient, and the preset learning rate, the model parameters corresponding to each layer of the initial abnormal type recognition model are updated to obtain an updated weight model, and based on the updated model parameters, an updated abnormal type recognition model is obtained.

[0013] In one embodiment, the method further comprises:

[0014] Collecting initial data to be detected, wherein the initial data to be detected includes amplitude information and phase information of the vibration signal;

[0015] Processing the initial data to be detected based on the data structure of the sample data to obtain target data to be detected, wherein the target data to be detected represents amplitude information and phase information of a vibration signal at a target position collected at a target time point;

[0016] The target data to be detected is predicted using the trained target anomaly type recognition model to obtain a predicted anomaly type corresponding to the target position at the time point.

[0017] In one embodiment, the method further comprises:

[0018] If it is determined that the preset damage condition is met based on the predicted abnormality type corresponding to the target position at the time point, an alarm message is output.

[0019] In one embodiment, the processing of the initial data to be detected based on the data structure of the sample data to obtain target data to be detected includes:

[0020] Performing truncation and normalization processing on the initial data to be detected to obtain processed data to be detected;

[0021] The initial data to be detected is processed based on the data structure of the sample data to obtain target data to be detected.

[0022] In a second aspect, the present application also provides an abnormality identification device, comprising:

[0023] An acquisition module, used to acquire training data, wherein the training data includes a plurality of sample data and a sample abnormality type corresponding to each of the sample data, wherein the sample data is vibration signal data and environmental feature data collected at a target position at a target time;

[0024] A prediction module, used to process the training data through an initial anomaly type recognition model to obtain a predicted anomaly type corresponding to the sample data;

[0025] The training module is used to train the initial abnormality type recognition model based on the predicted abnormality type and the sample abnormality type to obtain a trained target abnormality type recognition model.

[0026] In a third aspect, the present application further provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:

[0027] Acquire training data, wherein the training data includes a plurality of sample data and a sample abnormality type corresponding to each of the sample data, wherein the sample data is amplitude information and phase information of a vibration signal collected at a target position at a target time;

[0028] Processing the training data through an initial anomaly type recognition model to obtain a predicted anomaly type corresponding to the sample data;

[0029] Based on the predicted abnormality type and the sample abnormality type, the initial abnormality type recognition model is trained to obtain a trained target abnormality type recognition model.

[0030] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the following steps are implemented:

[0031] Acquire training data, wherein the training data includes a plurality of sample data and a sample abnormality type corresponding to each of the sample data, wherein the sample data is amplitude information and phase information of a vibration signal collected at a target position at a target time;

[0032] Processing the training data through an initial anomaly type recognition model to obtain a predicted anomaly type corresponding to the sample data;

[0033] Based on the predicted abnormality type and the sample abnormality type, the initial abnormality type recognition model is trained to obtain a trained target abnormality type recognition model.

[0034] In a fifth aspect, the present application further provides a computer program product, including a computer program, which implements the following steps when executed by a processor:

[0035] Acquire training data, wherein the training data includes a plurality of sample data and a sample abnormality type corresponding to each of the sample data, wherein the sample data is amplitude information and phase information of a vibration signal collected at a target position at a target time;

[0036] Processing the training data through an initial anomaly type recognition model to obtain a predicted anomaly type corresponding to the sample data;

[0037] Based on the predicted abnormality type and the sample abnormality type, the initial abnormality type recognition model is trained to obtain a trained target abnormality type recognition model.

[0038] The above-mentioned abnormality identification method, device, computer equipment, computer-readable storage medium and computer program product, wherein the method includes: obtaining training data, the training data including multiple sample data and sample abnormality types corresponding to each sample data, the sample data being the amplitude information and phase information of the vibration signal collected at the target position at the target time; processing the training data through an initial abnormality type identification model to obtain a predicted abnormality type corresponding to the sample data; training the initial abnormality type identification model based on the predicted abnormality type and the sample abnormality type to obtain a trained target abnormality type identification model.

[0039] By adopting this method, training can be performed based on multiple sample data and the sample anomaly types corresponding to each of the sample data to obtain a trained recognition model, thereby reducing training costs, improving computing energy efficiency, and reducing the complexity of model processing. Identification of abnormal situations based on the trained recognition model can also further improve the efficiency of network operation and maintenance, shorten the calculation time of external damage identification, and improve the ability to analyze and locate abnormal situations in optical fiber cables. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the drawings required for use in the embodiments of the present application or related technical descriptions will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0041] Figure 1 A schematic diagram of a flow chart of an abnormality identification method in one embodiment;

[0042] Figure 2 A schematic diagram of a flow chart of steps for obtaining a target anomaly type identification model in one embodiment;

[0043] Figure 3 A schematic diagram of a process for predicting anomaly types in one embodiment;

[0044] Figure 4 A schematic diagram of a flow chart of steps for obtaining target data to be detected in one embodiment;

[0045] Figure 5 A schematic diagram of a flow chart of an abnormality identification method in another embodiment;

[0046] Figure 6 A schematic diagram of a training process of a target anomaly type recognition model in an anomaly recognition method in an embodiment;

[0047] Figure 7 A schematic diagram of a training process of a target anomaly type recognition model in an anomaly recognition method in another embodiment;

[0048] Figure 8 In another embodiment of the abnormality identification method, Figure 5 A schematic diagram of an application flow of an abnormality identification method in another embodiment;

[0049] Fig. 9 is a structural block diagram of an abnormality identification device in an embodiment;

[0050] Fig.10 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0051] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0052] The abnormality identification method provided in the embodiment of the present application is as follows: Figure 1 As shown, a method for abnormal identification is provided, and the execution subject of the abnormal identification method can be a terminal device or a server or other processing device, wherein the terminal device can be a user equipment (User Equipment, UE), a mobile device, a user terminal, a terminal, a cellular phone, a cordless phone, a personal digital processing (Personal Digital Assistant, PDA), a handheld device, a computing device, a vehicle-mounted device, a wearable device, etc. In some possible implementations, the abnormal identification method can be implemented by a processor calling a computer-readable instruction stored in a memory. It can be understood that the method can also be applied to a system including a terminal and a server, and is implemented by the interaction of the terminal and the server. In this embodiment, the abnormal identification method is applied to a terminal device as an example for illustration, and the abnormal identification method includes the following steps:

[0053] Step 102: Obtain training data.

[0054] The training data includes a plurality of sample data and a sample anomaly type corresponding to each sample data, and the sample data is the amplitude information and phase information of the vibration signal collected at the target position at the target time. The vibration signal may be a Rayleigh scattering signal in the optical fiber, and the sample data may include a plurality of data pairs, and each data pair may be the amplitude information and phase information of the vibration signal at the target position collected at the target time, and the training data also includes a sample anomaly type corresponding to each data pair, and the sample anomaly type characterizes the abnormal condition of the optical fiber / optical cable at the target position at the target time, and the abnormal condition may include one or more of background noise, noise of mining mobile equipment, noise of cars, and noise of pedestrians.

[0055] Specifically, the terminal device can obtain multiple pieces of data to form training data. The terminal device can collect data through a box-type wavelength division φ-OTDR board card. For example, the box-type wavelength division φ-OTDR board card device and the Telemetry real-time data collection method can be used to collect the amplitude and phase information of the Rayleigh scattering signal at multiple locations and multiple times. The Rayleigh scattering signal can be caused by the vibration of the surrounding environment in the optical fiber; the type of the surrounding environment vibration can be an abnormal type, including one or more of background noise, noise of mining mobile equipment, noise of cars, and noise of pedestrians; in one example, the sample data can include the original data set A ij , A ij The data set generation process can be to collect M points in space and N points in time. Aij can be stored as an M×N two-dimensional array, including i types of external breaking patterns. Each type of external breaking pattern has j groups of data, that is, each group of data has a corresponding external breaking pattern, that is, the sample anomaly type.

[0056] Step 104: Process the training data using the initial anomaly type recognition model to obtain a predicted anomaly type corresponding to the sample data.

[0057] Among them, the initial abnormality type recognition model can be a model to be trained, for example, it can be a neural network model, which is used to identify the abnormal signal type corresponding to the vibration signal causing the target position, and the predicted abnormality type can be the abnormal signal type corresponding to the vibration signal causing the target position identified by the initial abnormality type recognition model after processing the sample data.

[0058] Specifically, the terminal device can train the initial abnormality type recognition model through training data. For example, multiple sample data and the sample abnormality type corresponding to each sample data can be input into the initial abnormality type recognition model. Each sample data and the sample abnormality type corresponding to each sample data can be processed by the initial abnormality type recognition model to obtain the predicted abnormality type corresponding to each sample data.

[0059] Step 106: Based on the predicted abnormality type and the sample abnormality type, the initial abnormality type recognition model is trained to obtain a trained target abnormality type recognition model.

[0060] Among them, the predicted anomaly type is the anomaly type corresponding to each sample data after processing each sample data through the initial anomaly type recognition model; the sample anomaly type is the anomaly type corresponding to each sample data obtained in advance in the actual situation.

[0061] Specifically, the terminal device can iteratively train the initial abnormality type recognition model through the predicted abnormality types corresponding to each sample data output by the initial abnormality type recognition model and the sample abnormality types corresponding to each sample data until the preset training completion conditions are met. The terminal device can obtain the trained abnormality type recognition model, that is, the target abnormality type recognition model.

[0062] In the above-mentioned abnormality identification method, training data is obtained, and the training data includes multiple sample data and sample abnormality types corresponding to each sample data. The sample data is the amplitude information and phase information of the vibration signal collected at the target position at the target time. The training data is processed by the initial abnormality type identification model to obtain the predicted abnormality type corresponding to the sample data. Based on the predicted abnormality type and the sample abnormality type, the initial abnormality type identification model is trained to obtain a trained target abnormality type identification model. By adopting this method, training can be performed based on multiple sample data and the sample abnormality types corresponding to each sample data to obtain a trained recognition model, thereby reducing training costs, improving computing energy efficiency, and reducing the complexity of model processing. Identification of abnormal situations based on the trained recognition model can also further improve the efficiency of network operation and maintenance, and can also shorten the calculation time of external damage identification, and can improve the analysis ability and abnormal situation location ability of optical fiber cables.

[0063] In an exemplary embodiment, Figure 2 As shown, the specific implementation process of the step "training the initial abnormality type recognition model based on the predicted abnormality type and the sample abnormality type to obtain the trained target abnormality type recognition model" may include:

[0064] Step 202: Obtain a loss value based on the loss function, the predicted anomaly type, and the sample anomaly type.

[0065] Among them, the loss function is used to calculate the loss value between the predicted anomaly type corresponding to each sample data and the sample anomaly type corresponding to each sample data. The loss value represents the difference between the predicted anomaly type output by the model corresponding to the sample data and the actual sample anomaly type.

[0066] Specifically, the terminal device can input the predicted anomaly type and sample anomaly type corresponding to each sample data into the loss function, calculate the predicted anomaly type and sample anomaly type corresponding to each sample data through the loss function, and obtain the loss value corresponding to each sample data respectively.

[0067] Step 204, when the preset training completion conditions are not met, the weight parameters of the initial abnormality type recognition model are updated based on the loss value to obtain an updated abnormality type recognition model, and based on the updated abnormality type recognition model, the step of obtaining training data is re-executed until the preset training completion conditions are met to obtain a trained target abnormality type recognition model.

[0068] Among them, the preset training completion condition can be that the calculated loss value has converged, or the training time reaches the preset time threshold, or the number of training iterations has reached the preset target number. The present disclosure does not limit the specific value of the preset time threshold and the specific value of the preset target number, and those skilled in the art can determine it specifically according to the actual application scenario. The initial abnormal type recognition model may include multiple computing layers; the weight parameters in the initial abnormal type recognition model may include the weight parameters in each computing layer in the initial abnormal type recognition model.

[0069] Specifically, the terminal device can input the training data into the initial abnormality type recognition model to obtain the predicted abnormality type corresponding to each sample data. If the terminal device determines that the current initial abnormality type recognition model does not meet the preset training completion conditions, the terminal device can calculate the loss value, and update the weight parameters of the initial abnormality type recognition model based on the loss value to obtain the updated initial abnormality type recognition model, and re-execute the step of obtaining the training data based on the updated initial abnormality type recognition model until the terminal device determines that the current initial abnormality type recognition model meets the preset training completion conditions. The terminal device can determine the current initial abnormality type recognition model as the trained abnormality type recognition model, that is, obtain the trained target abnormality type recognition model.

[0070] In this embodiment, by adjusting the initial abnormal type recognition model based on the training data and performing multiple rounds of iterative training, a model with better type recognition performance can be obtained, thereby ensuring a balance between the model training cost and the model type recognition performance.

[0071] In an exemplary embodiment, the specific implementation process of the step of "updating the weight parameters corresponding to each layer of the initial abnormal type recognition model based on the loss value to obtain an updated abnormal type recognition model" may include:

[0072] Based on the loss value, gradient, and preset learning rate, the model parameters corresponding to each layer of the initial abnormality type recognition model are updated to obtain updated model parameters, and an updated abnormality type recognition model is obtained based on the updated model parameters.

[0073] The initial abnormal type recognition model may include multiple computing layers, for example, the first layer, the second layer, ..., the nth layer, and the corresponding model parameters may include the model parameters of the first layer, the model parameters of the second layer, ..., the model parameters of the nth layer, etc. The specific value of the preset learning rate η may be determined based on the actual application scenario. The model parameters may include weight parameters and biases of each computing layer.

[0074] Specifically, the updating process of the model parameters of the initial abnormal type recognition model may include parameter updates in multiple training processes, and the updating process of the model parameters in each training process may be to update the weight parameters of each layer, for example, it may include: the initial abnormal type recognition model may include n computing layers, and the updating process of the model parameters of the target layer in the n computing layers may include that the terminal device may obtain the input data, output data, and the gradient of the return of the next layer of the target layer, and calculate the loss function based on the gradient of the return of the next layer of the target layer, the transposed data of the weight parameters of the next layer of the target layer, and the derivative of the excitation function of the input data of the target layer. Gradient of the input of the target layer; based on the gradient of the input of the target layer and the transposed data of the output data of the previous layer of the target layer, the gradient of the weight parameter of the target layer is calculated; based on the gradient of the input of the target layer, the gradient of the bias of the target layer is calculated; based on the gradient of the weight parameter of the target layer, the preset learning rate and the weight parameter of the target layer, the updated weight parameter of the target layer is updated; based on the gradient of the bias of the target layer, the preset learning rate and the bias of the target layer, the updated bias of the target layer is updated, and the terminal device can determine the updated weight parameter of the target layer and the updated bias of the target layer as the model parameters of the updated target layer. The terminal can repeat the above process until the updated model parameters of the nth layer are obtained, that is, the updated model parameters of each layer can be obtained, and the model parameters of each calculation layer in the initial abnormal type recognition model are respectively configured as the model parameters of each updated layer to obtain the updated initial abnormal recognition model, until the preset training completion condition is currently met, and the trained target abnormal type recognition model is obtained.

[0075] In an example, the target layer may be any computing layer in the initial anomaly type recognition model, and the process of updating the model parameters of the target layer may include:

[0076] The terminal device can calculate the gradient δ of the loss function for the input of the target layer by the following formula [l] :

[0077]

[0078] δ [;] =(Θ [;+1] ·(W [l+1]T ))⊙f′(z [l] )

[0079] Where L is the difference between the predicted anomaly type of the sample data in the initial anomaly type recognition model and the sample anomaly type; z [l] represents the input data of the target layer, Θ [l+1] represents the gradient of the next layer back to the target layer, f′(z [l] ) represents the derivative of the activation function of the input data of the target layer. [l+1]T Represents the transposed data of the weight parameters of the next layer of the target layer.

[0080] The terminal device can calculate the product of the transposed data of the weight parameters of the next layer of the target layer and the gradient returned by the next layer of the target layer, and perform XOR calculation on the product and the derivative of the excitation function of the input data of the target layer to obtain the gradient of the loss function for the input of the target layer.

[0081] The terminal device can calculate the gradient of the weight parameter of the target layer by the following formula

[0082]

[0083] Among them, a [l-1] Represents the output data of the previous layer of the target layer, a [l-1]T Represents the transposed data of the output data of the previous layer of the target layer; the terminal device can calculate the product of the gradient of the loss function with respect to the input of the target layer and the transposed data of the output data of the previous layer of the target layer, and determine the product as the gradient of the weight parameter of the target layer.

[0084] The terminal device can calculate the gradient of the target layer's bias using the following formula:

[0085]

[0086] Among them, b [l] Representing the bias of the target layer, the terminal device can determine the gradient of the loss function with respect to the input of the target layer as the gradient of the bias of the target layer.

[0087] The terminal device can calculate the updated weight parameter of the target layer by the following formula

[0088]

[0089] in, represents the weight parameter of the target layer before updating, and η represents the preset learning rate; Represents the gradient of the weight parameters of the target layer; the terminal device can calculate the product value between the preset learning rate and the gradient of the weight parameters of the target layer, and calculate the difference between the weight parameters of the target layer before updating and the product value, and determine the calculated difference as the weight parameters of the target layer after updating.

[0090] The terminal device can calculate the updated target layer bias by the following formula:

[0091]

[0092] in, represents the bias of the target layer before updating, Represents the gradient of the bias of the target layer; the terminal device can calculate the product of the preset learning rate and the gradient of the bias of the target layer, and calculate the difference between the bias of the target layer before updating and the product value, and determine the calculated difference as the bias of the target layer after updating.

[0093] Based on the above, the terminal device can determine the updated weight parameters of the target layer and the updated bias of the target layer as the model parameters of the updated target layer. The terminal can repeat the above process until the updated model parameters of each layer are obtained, and the model parameters of each calculation layer in the initial abnormality type recognition model are respectively configured as the model parameters of each updated layer to obtain the updated initial abnormality recognition model, until the preset training completion conditions are met, and the trained target abnormality type recognition model is obtained.

[0094] In this embodiment, the preset learning rate can flexibly control the step size of weight update, ensure the accuracy and effect of training, and improve the performance of model training.

[0095] In an exemplary embodiment, Figure 3 As shown, the abnormality identification method also includes:

[0096] Step 302: Collect initial data to be detected.

[0097] Specifically, the initial data to be detected includes the amplitude information and phase information of the vibration signal. The initial data to be detected may be the amplitude information and phase information of the vibration signal collected at the position to be detected. The initial data to be detected may also be the amplitude information and phase information of the vibration signal collected at the position to be detected at the time to be detected.

[0098] Step 304: Process the initial data to be detected based on the data structure of the sample data to obtain target data to be detected.

[0099] The target data to be detected represents the amplitude information and phase information of the vibration signal at the target position collected at the target time point, and the data structure of the sample data can be the data structure of the input data of the target abnormality type recognition model.

[0100] Specifically, the terminal device can collect initial data to be detected at the location to be detected at the time to be detected, and adjust the data structure of the initial data to be detected based on the data structure of the input data of the target abnormality type recognition model. For example, the data structure of the initial data to be detected can be converted into the data structure of the input data of the target abnormality type recognition model to obtain the target data to be detected. The target data to be detected can be {Ck, L}, where L can represent the collection location information of the target data to be detected.

[0101] Step 306 , predict the target data to be detected by using the trained target anomaly type recognition model to obtain the predicted anomaly type corresponding to the target position at the time point.

[0102] Specifically, the terminal device can input the target data to be detected into the target abnormality type recognition model, perform feature analysis on the target data to be detected through the target abnormality type recognition model, and output the predicted abnormality type corresponding to the target data to be detected, that is, the external damage situation corresponding to the target position point identified by the target abnormality type recognition model at the current time point. The output result of the model can be the abnormal signal type corresponding to the vibration signal that causes the target position point identified by the model.

[0103] In this embodiment, a binary neural network model is used to realize online identification and monitoring of abnormal conditions of optical cables that can be quickly determined, and the traditional neural network model is quantified, which effectively reduces the memory space occupied by the model and the number of operations of the model output processing results, avoids manual fault analysis of the vibration signal of the optical fiber, and can monitor and predict abnormal conditions of optical fiber cables in real time, which is convenient for system operation and maintenance personnel to discover and troubleshoot in time, thereby improving the stability and reliability of the system.

[0104] In an exemplary embodiment, the abnormality identification method further includes:

[0105] If the preset damage condition is determined to be met based on the predicted anomaly type corresponding to the target location at the time point, an alarm message is output.

[0106] The content of the preset damage condition may be that the predicted abnormality type is the target abnormality type, and the specific content of the target abnormality type may be determined based on the actual application scenario. In one example, the target abnormality type may be the noise of the mining mobile device. In another example, the preset damage condition may be that the predicted abnormality type corresponding to the target location point will cause damage to the optical fiber cable.

[0107] Specifically, the terminal device can process the data to be detected through the target anomaly type recognition model to obtain the predicted anomaly type corresponding to the data to be detected. The data to be detected can be collected at the target position and target time point. If it is determined that the current predicted anomaly type meets the preset damage condition, the terminal needs to output alarm information, which needs to include at least the target position point corresponding to the predicted anomaly type and the predicted anomaly type.

[0108] In one example, if the terminal device determines that the currently predicted abnormal type is noise from the mining mobile device, it can be determined that the preset damage condition is met. The terminal device can output the alarm information to the operation and maintenance personnel so that the operation and maintenance personnel can intervene.

[0109] In this embodiment, abnormal conditions of optical fiber cables can be monitored in real time, and alarm information can be output to operation and maintenance personnel in a timely manner, so as to facilitate timely maintenance of the optical fiber and ensure the stability of the communication system.

[0110] In an exemplary embodiment, Figure 4 As shown, the specific implementation process of the step of "processing the initial data to be detected based on the data structure of the sample data to obtain the target data to be detected" may include:

[0111] Step 402 , truncate and normalize the initial data to be detected to obtain processed data to be detected.

[0112] The initial data to be detected may be data obtained after detecting multiple locations at the same time point.

[0113] Specifically, the terminal device can truncate the initial data to be detected according to the position coordinate information of each position point to obtain the data to be detected corresponding to each position point. For the data to be detected corresponding to each position point, the terminal device can normalize the data to be detected at each position point to obtain the processed data to be detected corresponding to each position point.

[0114] Step 404 , processing the processed data to be detected based on the data structure of the sample data to obtain target data to be detected.

[0115] The processed data to be detected include the processed data to be detected corresponding to each position point.

[0116] Specifically, for the data to be detected corresponding to each position point, the terminal device can adjust the data structure of the data to be detected corresponding to each position point based on the data structure of the input data of the target abnormality type recognition model. For example, the data structure of the data to be detected corresponding to each position point can be converted into the data structure of the input data of the target abnormality type recognition model to obtain the target data to be detected.

[0117] In this embodiment, the collected raw data may be processed to ensure data normalization, thereby facilitating the use of the target anomaly type recognition model.

[0118] In the following, in conjunction with a specific embodiment, a specific implementation process of the above-mentioned abnormality identification method is described in detail, which may include:

[0119] In view of the risk of optical cables being routed together due to large-scale expansion, splicing, missing labels, etc., and the external damage of optical cables caused by human damage such as construction, intelligent models can be used to manage dumb resources of optical networks, reduce operation and maintenance costs, and improve the level of intelligence. However, the learning cost and complexity of the external damage recognition algorithm of the existing intelligent model are relatively high. The abnormal type recognition model in the abnormal recognition method provided by the present disclosure can effectively reduce the hardware implementation complexity and training cost of the recognition algorithm, improve computing energy efficiency, and is of great significance to improving network operation and maintenance efficiency.

[0120] The abnormality identification method provided by the present disclosure can be an optical fiber cable external damage identification method based on a binary neural network. Through real-time data collection by Telemetry and analysis of the amplitude information and phase information of the Rayleigh scattering signal in the optical fiber cable collected by the box-type wavelength division φ-OTDR board, a binary artificial neural network is used to identify the external damage of the optical fiber cable. The computational complexity of the network can be reduced by binarizing the artificial neural network, shortening the calculation time of external damage identification, so that it can quickly and accurately analyze and predict reliability before the optical fiber cable fails, thereby greatly improving the ability of optical fiber cable fault analysis and fault location.

[0121] Specifically, Figure 5 As shown, it may be a flowchart of the abnormality identification method provided by the present disclosure:

[0122] Collect optical fiber cable related data, pre-process the data, divide it into training set and test set, train and optimize parameters based on the training set and binary neural network algorithm, and obtain a binary neural network model (i.e., target anomaly type recognition model). The binary neural network model is used to recognize external damage patterns (anomaly types).

[0123] Specifically, Figure 6As shown, it may be a flowchart of the training process of the target abnormality type recognition model in the abnormality recognition method provided by the present disclosure:

[0124] Step 1: Collect the original data {Aij} of the optical fiber cable; the box-type wavelength division φ-OTDR board uses Telemetry technology to collect the amplitude and phase information of the Rayleigh scattering signal in the optical fiber caused by the vibration of the external surrounding environment in real time as the original data set Aij, collect M points in space and N points in time, then A ij It can be stored as a two-dimensional array of M×N, including i types of external breaking patterns, and each type of external breaking pattern has j groups of data.

[0125] Step 2: truncate and group the data, and normalize and preprocess the data {B i,j}, set the label {Y i Specifically, the terminal device groups and normalizes the original data set to obtain the target data set B i,j , the target data set B can be obtained by the following formula i,j :

[0126] B i,j =(A i,j –min(A i,j )) / max(A i,j )-min(A i,j ))

[0127] The terminal device can process the preprocessed target data set B i,j The labels of the abnormal types corresponding to the data pairs (i.e., sample data) contained in the target data set are obtained by associating them with the external damage of the surrounding optical fiber cables (abnormal types), which can be defined as {Y i}.

[0128] Step 3: Network parameters (Nl, N w etc.) settings, train the network model ρ; specifically, configure the relevant binary network parameters, including the number of layers N l , the number of weights N w Etc., use the prepared data set to train the model (initial anomaly type recognition model).

[0129] In one example, the model parameter updating process may further include: the target layer may be any computing layer in the initial anomaly type recognition model, and the model parameter updating process for the target layer may include:

[0130] The terminal device can calculate the gradient δ of the loss function for the input of the target layer by the following formula [l] :

[0131]

[0132] δ [l] =(Θ [l+1] ·(W [l+1]T ))⊙f′(z [l] )

[0133] Where L is the difference between the predicted anomaly type of the sample data in the initial anomaly type recognition model and the sample anomaly type; z [l] represents the input data of the target layer, Θ [l+1] represents the gradient of the next layer back to the target layer, f′(z [l] ) represents the derivative of the activation function of the input data of the target layer. [l+1]T Represents the transposed data of the weight parameters of the next layer of the target layer.

[0134] The terminal device can calculate the product of the transposed data of the weight parameters of the next layer of the target layer and the gradient returned by the next layer of the target layer, and perform XOR calculation on the product and the derivative of the excitation function of the input data of the target layer to obtain the gradient of the loss function for the input of the target layer.

[0135] The terminal device can calculate the gradient of the weight parameter of the target layer by the following formula

[0136]

[0137] Among them, a [l-1] Represents the output data of the previous layer of the target layer, a [l-1]T Represents the transposed data of the output data of the previous layer of the target layer; the terminal device can calculate the product of the gradient of the loss function with respect to the input of the target layer and the transposed data of the output data of the previous layer of the target layer, and determine the product as the gradient of the weight parameter of the target layer.

[0138] The terminal device can calculate the gradient of the target layer's bias using the following formula:

[0139]

[0140] Among them, b [l] Representing the bias of the target layer, the terminal device can determine the gradient of the loss function with respect to the input of the target layer as the gradient of the bias of the target layer.

[0141] The terminal device can calculate the updated weight parameter of the target layer by the following formula

[0142]

[0143] in, represents the weight parameter of the target layer before updating, and η represents the preset learning rate; Represents the gradient of the weight parameters of the target layer; the terminal device can calculate the product value between the preset learning rate and the gradient of the weight parameters of the target layer, and calculate the difference between the weight parameters of the target layer before updating and the product value, and determine the calculated difference as the weight parameters of the target layer after updating.

[0144] The terminal device can calculate the updated target layer bias by the following formula:

[0145]

[0146] in, represents the bias of the target layer before updating, Represents the gradient of the bias of the target layer; the terminal device can calculate the product of the preset learning rate and the gradient of the bias of the target layer, and calculate the difference between the bias of the target layer before updating and the product value, and determine the calculated difference as the bias of the target layer after updating.

[0147] Based on the above, the terminal device can determine the updated weight parameters of the target layer and the updated bias of the target layer as the model parameters of the updated target layer. The terminal can repeat the above process until the updated model parameters of each layer are obtained, and the model parameters of each calculation layer in the initial abnormality type recognition model are respectively configured as the model parameters of each updated layer to obtain the updated initial abnormality recognition model, until the preset training completion conditions are met, and the trained target abnormality type recognition model is obtained.

[0148] Step 4: Save the network model ρ. Specifically, iteratively optimize the network parameters according to the training result feedback to meet the accuracy requirements. Save the binary network model ρ.

[0149] like Figure 7As shown, it can be a schematic diagram of the training process of the target anomaly type recognition model. The input data can be a space-time sampling matrix (10000*12*1). The input data is input to the first convolutional layer, and the size of the first convolutional layer is (200*3)*1*5. The output result of the first convolutional layer is obtained. The output result of the first convolutional layer is input to the second convolutional layer for binary quantization processing to obtain the output result of the second convolutional layer. The output result of the second convolutional layer is input to the third convolutional layer for binary quantization processing to obtain the output result of the third convolutional layer. The size of the second convolutional layer is different from that of the third convolutional layer. The second convolutional layer The size of can be larger than the size of the third convolutional layer; the output result of the third convolutional layer is input into the maximum pooling layer to obtain the maximum pooling result, the maximum pooling result is input into the fully connected layer to obtain the output result of the model, and the loss value (Loss) corresponding to the loss function is calculated based on the model output result and the label, and the update is performed based on the loss value corresponding to the loss function to obtain the updated model parameters w, b, and the initial anomaly type recognition model is configured based on the updated model parameter number w, b to obtain the updated anomaly type recognition model, until the preset training completion condition is met, and the trained target anomaly type recognition model is obtained.

[0150] like Figure 8 As shown, this can be the application process of the target anomaly type recognition model:

[0151] Collecting optical fiber cable related data at position L, that is, collecting amplitude information and phase information of the vibration signal at position L;

[0152] The data is truncated and normalized to obtain {C k , L};

[0153] Load the network model ρ, that is, load the target anomaly type recognition model ρ;

[0154] External damage identification {Dk}, that is, the target abnormality type identification model processes the amplitude information and phase information of the vibration signal at position L, and identifies the external damage corresponding to position L, that is, the abnormality type;

[0155] Determine whether reporting is necessary. If reporting is necessary, report the external damage and location information to the network management.

[0156] In one example, the deployed φ-OTDR equipment is used to collect data related to the external environment around the optical fiber cable and input it into the algorithm model, and the external damage situation can be output after calculation. The steps are as follows:

[0157] The data is truncated and normalized, and matched with the saved network input data structure to obtain {C k,L}, L represents the location information. Load the previously saved algorithm model ρ, input the data collected at the current time point into the algorithm model, and the model will output the current external damage situation {Dk,L}=ρ(Ck,L). External damage situation {D k}Including background noise D1, excavation D2, cars D3, pedestrians D4, etc., the recognition results and location information L are reported to the network management system. The network management makes a judgment based on the current recognition situation. If the optical fiber cable is damaged, the network management will issue an alarm message and manual intervention will be carried out in advance.

[0158] The present application provides an abnormal identification method that can realize the identification of optical cable channel construction threat events based on the Φ-OTDR of the neural network; the binary neural network is used to binarize the weights of the traditional neural network, so as to solve the problems of high computational complexity and inconvenience of hardware implementation of the traditional neural network algorithm, and improve the recognition and computational efficiency of the optical fiber cable external break mode. Through binary calculation, all calculations of the huge optical cable vibration data are transformed into bit operations, and the memory occupied is 1 / 32 of the traditional NN. The energy consumed during the multiplication operation is significantly reduced, which is more conducive to the implementation of the hardware chip. The binary artificial neural network technology is used to realize the online monitoring of the external break of the optical cable in seconds. The binary artificial neural network is a quantification of the traditional neural network, which effectively reduces the memory usage and the number of floating point operations. Combined with elemetry real-time data collection and analysis of the box-type wavelength division φ-OTDR board, there is no need to manually locate the fault. Relying on the second-level reporting of performance data, the external break can be predicted in real time, and a report can be provided to the network management system for use by operation and maintenance personnel. The analysis and prediction of the external break of the optical fiber cable are realized, and the reliability of the box-type wavelength division equipment is further improved. When the management and control system is connected to equipment from multiple manufacturers, it can adapt to optoelectronic equipment from the same manufacturer and optical equipment from the same manufacturer to electrical equipment from other manufacturers.

[0159] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0160] Based on the same inventive concept, the embodiment of the present application also provides an abnormality identification device for implementing the abnormality identification method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more abnormality identification device embodiments provided below can refer to the limitations on the abnormality identification method above, and will not be repeated here.

[0161] In an exemplary embodiment, Fig. 9 As shown, an abnormality identification device 900 is provided, comprising:

[0162] An acquisition module 902 is used to acquire training data, wherein the training data includes a plurality of sample data and a sample abnormality type corresponding to each of the sample data, wherein the sample data is vibration signal data and environmental feature data collected at a target position at a target time;

[0163] Prediction module 904, used to process the training data through the initial abnormality type recognition model to obtain the predicted abnormality type corresponding to the sample data;

[0164] The training module 906 is used to train the initial abnormality type recognition model based on the predicted abnormality type and the sample abnormality type to obtain a trained target abnormality type recognition model.

[0165] Each module in the above-mentioned abnormality identification device can be implemented in whole or in part by software, hardware and a combination thereof. Each of the above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory in the computer device in the form of software, so that the processor can call and execute the corresponding operations of each of the above modules.

[0166] In an exemplary embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as shown in FIG. Fig.10As shown. The computer device includes a processor, a memory, an input / output interface, a communication interface, a display unit and an input device. Among them, the processor, the memory and the input / output interface are connected through a system bus, and the communication interface, the display unit and the input device are connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and the external device. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be implemented through WIFI, a mobile cellular network, near field communication (Near Field Communication, NFC) or other technologies. When the computer program is executed by the processor, an abnormality identification method is implemented. The display unit of the computer device is used to form a visually visible picture, which can be a display screen, a projection device or a virtual reality imaging device. The display screen can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covering the display screen, or a button, trackball or touchpad set on the computer device shell, or an external keyboard, touchpad or mouse.

[0167] Those skilled in the art will understand that Fig.10 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0168] In an exemplary embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps described in the above embodiment when executing the computer program.

[0169] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps described in the above embodiment are implemented.

[0170] In one embodiment, a computer program product is provided, including a computer program, which implements the steps described in the above embodiment when executed by a processor.

[0171] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0172] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, an artificial intelligence (AI) processor, etc., but are not limited to this.

[0173] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0174] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A method for identifying anomalies, characterized in that: The method comprises: Acquire training data, wherein the training data includes a plurality of sample data and a sample abnormality type corresponding to each of the sample data, wherein the sample data is amplitude information and phase information of a vibration signal collected at a target position at a target time; Processing the training data through an initial anomaly type recognition model to obtain a predicted anomaly type corresponding to the sample data; Based on the predicted abnormality type and the sample abnormality type, the initial abnormality type recognition model is trained to obtain a trained target abnormality type recognition model.

2. The method according to claim 1, characterized in that The initial abnormality type recognition model is trained based on the predicted abnormality type and the sample abnormality type to obtain a trained target abnormality type recognition model, including: Obtaining a loss value based on the loss function, the predicted anomaly type, and the sample anomaly type; When the preset training completion conditions are not met, the weight parameters of the initial abnormality type recognition model are updated based on the loss value to obtain an updated abnormality type recognition model, and based on the updated abnormality type recognition model, the step of obtaining training data is re-executed until the preset training completion conditions are met to obtain a trained target abnormality type recognition model.

3. The method according to claim 2, characterized in that The updating of the weight parameters corresponding to each layer of the initial abnormal type recognition model based on the loss value to obtain an updated abnormal type recognition model includes: Based on the loss value, the gradient, and the preset learning rate, the model parameters corresponding to each layer of the initial abnormal type recognition model are updated to obtain an updated weight model, and based on the updated model parameters, an updated abnormal type recognition model is obtained.

4. The method according to claim 1, characterized in that: The method further comprises: Collecting initial data to be detected, wherein the initial data to be detected includes amplitude information and phase information of the vibration signal; Processing the initial data to be detected based on the data structure of the sample data to obtain target data to be detected, wherein the target data to be detected represents amplitude information and phase information of a vibration signal at a target position collected at a target time point; The target data to be detected is predicted using the trained target anomaly type recognition model to obtain a predicted anomaly type corresponding to the target position at the time point.

5. The method according to claim 4, characterized in that The method further comprises: If it is determined that the preset damage condition is met based on the predicted abnormality type corresponding to the target position at the time point, an alarm message is output.

6. The method according to claim 4, characterized in that The processing of the initial data to be detected based on the data structure of the sample data to obtain target data to be detected includes: Performing truncation and normalization processing on the initial data to be detected to obtain processed data to be detected; The processed data to be detected is processed based on the data structure of the sample data to obtain target data to be detected.

7. An abnormality identification device, characterized in that: The device comprises: An acquisition module, used to acquire training data, wherein the training data includes a plurality of sample data and a sample abnormality type corresponding to each of the sample data, wherein the sample data is vibration signal data and environmental feature data collected at a target position at a target time; A prediction module, used to process the training data through an initial anomaly type recognition model to obtain a predicted anomaly type corresponding to the sample data; The training module is used to train the initial abnormality type recognition model based on the predicted abnormality type and the sample abnormality type to obtain a trained target abnormality type recognition model.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images