Optical network soft fault detection method and device, storage medium and electronic device

Abstract

The invention relates to the technical field of computers, and provides an optical network soft fault detection method and device, a storage medium and an electronic device, and the optical network soft fault detection method comprises the steps: obtaining to-be-detected signal data of a target optical network; and inputting the to-be-detected signal data to a pre-trained fault detection model to obtain fault detection information output by the fault detection model, the fault detection model being obtained by performing self-supervision training based on the unlabeled sample signal data. According to the optical network soft fault detection method, the detection efficiency of soft fault detection in the optical network is improved.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to an optical network soft fault detection method, device, storage medium and electronic device. Background Art

[0002] With the rapid development of high-speed, bandwidth-intensive services, the surge in data traffic has put forward higher requirements for the stability and flexibility of optical networks. As a key infrastructure for modern communications, the efficient operation and reliability of optical networks are directly related to the quality and continuity of communication services. In this context, ensuring the stability and reliability of optical networks has become an important issue that needs to be addressed urgently.

[0003] In optical networks, faults can be divided into hard faults and soft faults. Hard faults usually cause physical damage to the device or link and are easy to detect and repair; soft faults are different. They do not directly cause physical damage, but will cause signal quality degradation, which in turn affects communication performance. If not diagnosed and handled in time, soft faults may gradually evolve into serious link failures, causing greater losses. Therefore, early detection and diagnosis of soft faults are crucial to maintaining the stability of optical networks.

[0004] Traditional soft fault diagnosis methods mainly rely on setting thresholds to trigger alarms. These methods require a large amount of data and manual annotation to determine a reasonable threshold. However, if the threshold is set too high, a large number of missed reports will be generated, increasing the risk of business interruption; if it is set too low, a large number of false alarms will be generated, increasing the workload of operation and maintenance personnel. Summary of the invention

[0005] The present invention provides an optical network soft fault detection method, device, storage medium and electronic device, which are used to solve the problem that the traditional optical network soft fault detection process in the prior art requires a large amount of data and manual annotation to determine a reasonable threshold, thereby realizing a more efficient optical network soft fault detection process.

[0006] The present invention provides an optical network soft fault detection method, comprising obtaining signal data to be detected of a target optical network; inputting the signal data to be detected into a pre-trained fault detection model to obtain fault detection information output by the fault detection model, wherein the fault detection model is obtained by self-supervised training based on unlabeled sample signal data.

[0007] According to a method for detecting soft faults in an optical network provided by the present invention, the step of training the fault detection model includes: obtaining unlabeled sample signal data; calculating a first signal feature of the unlabeled sample signal data according to a feature index, wherein the feature index is selected based on prior knowledge; extracting a second signal feature from the unlabeled sample signal data based on an encoder; fusing the first signal feature and the second signal feature to obtain a pseudo-label for the unlabeled sample signal data; and training the initial network model based on the unlabeled sample signal data and the pseudo-label to obtain the fault detection model.

[0008] According to an optical network soft fault detection method provided by the present invention, it is characterized in that the encoder-based extraction of the second signal feature from the unlabeled sample signal data includes: using a convolutional autoencoder to automatically learn and extract the second signal feature from the unlabeled sample signal data.

[0009] According to an optical network soft fault detection method provided by the present invention, it is characterized in that the initial network model is trained based on the unlabeled sample signal data and the pseudo-label to obtain the fault detection model, including: performing self-supervised training on the initial network model based on the unlabeled sample signal data and the pseudo-label to obtain the initial fault detection model; obtaining labeled sample signal data and label data corresponding to the labeled sample signal data; and performing supervised fine-tuning on the initial fault detection model based on the labeled sample signal data and the label data to obtain the fault detection model.

[0010] According to a method for detecting soft faults in an optical network provided by the present invention, it is characterized in that the supervised fine-tuning of the initial fault detection model based on the labeled sample signal data and the label data to obtain the fault detection model includes: using a cross entropy loss function to evaluate the difference between the predicted label of the labeled sample signal data and the label data corresponding to the labeled sample signal data, and updating the weights of the initial fault detection model through a back propagation algorithm to minimize the value of the loss function.

[0011] According to an optical network soft fault detection method provided by the present invention, it is characterized in that the initial network model is established based on a deep convolutional neural network, and the deep convolutional neural network includes a preset number of convolutional layers, a preset number of pooling layers and a fully connected layer, each convolutional layer is followed by a pooling layer, and finally the prediction result is output through the fully connected layer.

[0012] According to an optical network soft fault detection method provided by the present invention, it is characterized in that the characteristic indicators include at least one of the following: information theory characteristic indicators, statistical description characteristic indicators, morphological characteristic indicators and ratio index characteristic indicators; the information theory characteristic indicators include information entropy; the statistical description characteristic indicators include at least one of the following: mean, variance, standard deviation, kurtosis and skewness; the morphological characteristic indicators include at least one of the following: maximum value, minimum value, full-width half-peak value and symmetry; the ratio index characteristic indicators include at least one of the following: waveform index, pulse index and peak index.

[0013] The present invention also provides an optical network soft fault detection device, comprising: an acquisition module, configured to acquire signal data to be detected of a target optical network; a detection module, configured to input the signal data to be detected into a pre-trained fault detection model to obtain fault detection information output by the fault detection model, wherein the fault detection model is obtained by self-supervised training based on unlabeled sample signal data.

[0014] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, any one of the optical network soft fault detection methods described above is implemented.

[0015] The present invention also provides a non-transitory computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the optical network soft fault detection method as described above is implemented.

[0016] The present invention also provides a computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the computer program implements any one of the optical network soft fault detection methods described above.

[0017] The optical network soft fault detection method, device, storage medium and electronic device provided by the present invention train a fault detection model based on a self-supervised training method, complete the training of the fault detection model when the training data is limited, and improve the detection efficiency of soft fault detection in the optical network. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0019] Figure 1 It is a flow chart of the optical network soft fault detection method provided by the present invention.

[0020] Figure 2 It is a schematic diagram of the process of training a fault detection model in the optical network soft fault detection method provided by the present invention.

[0021] Figure 3 It is a structural diagram of an exemplary fault detection model in the present invention.

[0022] Figure 4 It is a schematic diagram of the structure of an exemplary experimental system in the present invention.

[0023] Figure 5 It is a schematic diagram of the model performance experimental results of performing small sample soft fault detection using the first signal feature, the second signal, and the fusion feature of the first signal feature and the second signal in the present invention.

[0024] Figure 6 It is a schematic diagram of the model performance experimental results of embedding prior knowledge using different self-supervised learning models under small sample conditions in the present invention.

[0025] Figure 7 It is a structural schematic diagram of the optical network soft fault detection device provided by the present invention.

[0026] Figure 8 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0027] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0028] Combine the following Figure 1-Figure 8 The invention describes an optical network soft fault detection method, device, storage medium and electronic device.

[0029] Figure 1 FIG. 1 is a flow chart of a method for detecting soft faults in an optical network according to an embodiment of the present invention. Figure 1 As shown, the method includes: Step 101: Acquire signal data to be detected of a target optical network.

[0030] In this embodiment, the target optical network may be an optical network to be detected for soft faults. The signal data to be detected may be acquired by a receiver, a photodetector or other acquisition equipment, and the signal data to be detected may be a digital spectrum of the signal, such as amplitude spectrum density data obtained based on fast Fourier transform, or power spectrum density data obtained according to the periodogram method or the Welch method (a modified periodogram power spectrum density estimation method); or other signal data related to fault detection.

[0031] Step 102: input the signal data to be detected into a pre-trained fault detection model to obtain fault detection information output by the fault detection model.

[0032] In this embodiment, the fault detection model is used to characterize the correspondence between signal data and fault detection information. The fault detection information may include indication information indicating whether there is a soft fault in the optical network, and may also include the fault type and probability of the soft fault. The fault detection model is obtained by self-supervised training based on unlabeled sample signal data. The training of the fault detection model is an iterative and multi-stage process. First, it is necessary to collect and preprocess the unlabeled sample signal data. Then, features are automatically extracted by technologies such as autoencoders, and pseudo labels are generated by methods such as clustering algorithms. Using these pseudo labels, a deep learning model such as a deep convolutional neural network is constructed to learn the basic features of the data in the self-supervised pre-training stage. During the training process, a suitable loss function is defined and the model weights are updated accordingly. At the same time, the performance can be optimized and overfitting can be prevented by adjusting hyperparameters and applying regularization techniques. The model can then be evaluated on the validation set and its accuracy can be improved through iterative optimization. Finally, the tested and verified model can be deployed in practical applications for real-time fault detection, and its performance is continuously monitored for necessary maintenance and updates.

[0033] The optical network soft fault detection method provided by the present invention trains a fault detection model based on a self-supervised training method, completes the training of the fault detection model under the condition of limited training data, and improves the detection efficiency of soft fault detection in the optical network.

[0034] In some optional implementations, the process of fault detection model training in the optical network soft fault detection method can be found in Figure 2 ,include: Step 201: Obtain unlabeled sample signal data.

[0035] In this implementation, unlabeled sample signal data refers to data that does not provide any category or label information during data collection or processing. Such data is usually raw, unprocessed, or collected without any additional information about its content.

[0036] Step 202: Calculate the first signal feature of the unlabeled sample signal data according to the feature index.

[0037] In this implementation, feature indicators are selected based on prior knowledge. Prior knowledge refers to the knowledge or information about the problem that is already possessed before learning or solving the problem. This knowledge can be explicit or implicit, and usually comes from expert experience, theories in the field, statistical data, historical information, or previous learning results. In the field of optical network fault diagnosis, experts have accumulated a lot of empirical knowledge, which is crucial for building effective diagnostic models. By introducing additional information and constraints with the assistance of domain expert knowledge, the model can be guided to explore the data space more effectively during the learning process, which can provide more background information to alleviate data sparsity.

[0038] Step 203: extract a second signal feature from the unlabeled sample signal data based on the encoder.

[0039] In this implementation, the encoder can be established based on a convolutional neural network and can be trained through a supervised training method, but the supervised training method requires a certain amount of labeled sample data. Preferably, autoencoders such as convolutional autoencoders can be used to learn from the original unlabeled data and extract useful feature representations.

[0040] Step 204: The first signal feature and the second signal feature are combined to obtain a pseudo label for the unlabeled sample signal data.

[0041] In this implementation, the fusion of the first signal feature and the second signal feature can be a simple data concatenation, or a more complex weighted combination, feature interpolation, etc. Pseudo label is a concept in self-supervised learning methods. Self-supervised learning methods are particularly suitable for situations where labels are scarce. They can reduce the reliance on a large amount of manually annotated data and help the model learn more robust feature representations.

[0042] Step 205: Train the initial network model based on the unlabeled sample signal data and the pseudo labels to obtain a fault detection model.

[0043] In this implementation, the initial network model can select a suitable deep learning model architecture, such as a convolutional neural network (CNN), a recurrent neural network (RNN), a multilayer perceptron (MLP) or a transformer model (Transformer), depending on the characteristics of the signal and the requirements of the fault detection task.

[0044] This implementation method selects characteristic indicators according to prior knowledge to calculate signal characteristics, and performs self-supervisory training on the model based on the calculated signal characteristics, thereby improving the performance of the fault detection model and thus improving the detection efficiency of soft fault detection in optical networks.

[0045] In some optional implementations, extracting the second signal feature from the unlabeled sample signal data based on the encoder includes: using a convolutional autoencoder to automatically learn and extract the second signal feature from the unlabeled sample signal data. Extracting the second signal feature based on the convolutional autoencoder does not require supervised training of the encoder, further reducing the use of labeled samples, and this process does not require manual intervention, thereby improving the automation level of feature extraction.

[0046] Convolutional Autoencoder (CAE) has powerful feature learning capabilities. It can automatically learn and extract representative features directly from raw data without labels and manual intervention. Convolutional Autoencoder consists of an encoder and a decoder, where the encoder and decoder are composed of convolutional layers and transposed convolutional layers respectively. When using convolutional autoencoder for feature extraction, the data is first passed to the encoder, and then passes through two convolutional layers for feature extraction and dimension compression. After each convolutional layer, the rectified linear unit activation function is used to introduce nonlinearity, allowing the model to learn more complex feature representations. After processing the two convolutional layers, the data is flattened and then passed through the fully connected layer to obtain the extracted features. The role of the decoder is to reconstruct the original data through the extracted features. The quality of the reconstruction is evaluated by the loss function. As an example, the loss function used can be the mean squared error (MSE).

[0047] In some optional implementations, the initial network model is based on a deep convolutional neural network, which includes a preset number of convolutional layers, a preset number of pooling layers, and a fully connected layer. Each convolutional layer is followed by a pooling layer, and finally the prediction result is output through the fully connected layer. As an example, see Figure 3 , Figure 3 is a schematic diagram of the structure of an exemplary fault detection model in the present invention, Figure 3 The number of convolutional layers and pooling layers is 4. The specific number in actual use can be adjusted according to the training situation.

[0048] In some optional implementations, the initial network model is trained based on the unlabeled sample signal data and the pseudo-label to obtain the fault detection model, including: performing self-supervisory training on the initial network model based on the unlabeled sample signal data and the pseudo-label to obtain the initial fault detection model; obtaining the labeled sample signal data and the label data corresponding to the labeled sample signal data; and performing supervised fine-tuning on the initial fault detection model based on the labeled sample signal data and the label data to obtain the fault detection model. The performance of the model can be further improved by performing supervised fine-tuning, thereby improving the efficiency of fault detection.

[0049] In this implementation, the labeled sample signal data can be obtained through experiments or simulation experiments, such as collecting experimental data through a communication link built on the VPI photonic simulation platform, see Figure 4 , Figure 4 It is a schematic diagram of the structure of an exemplary experimental system in the present invention. Figure 4 A single-channel 30 GBaud PDM-QPSK link is built with a roll-off factor of 0.05 and a laser center frequency of 193.1 THz.

[0050] Figure 4 The experimental system may include the following components: Offline TX-DSP (offline transmit digital signal processing), responsible for offline processing of digital signals at the transmitter; DAC (digital-to-analog converter), which converts the digital signals processed by the offline TX-DSP into analog signals; I / Q modulator (I / Q modulator), which performs I / Q modulation on the analog signals and divides them into two paths; PBS (polarization beam splitter), which is used to separate two orthogonal polarization signals after I / Q modulation; PBC (polarization maintaining circuit), which maintains the signal in one polarization direction and allows the other polarization direction to pass; EDFA (erbium-doped fiber amplifier), which amplifies the signal by using stimulated radiation generated by erbium ions in erbium-doped fiber under laser excitation within a specific wavelength range; BV-WSS (broadband variable waveplate attenuator), which is used to adjust the attenuation of the optical path; WSS (waveplate switch system), which is used to switch different optical path configurations; CoRx (cooperative receiver), which may include photodiodes and amplifiers; ADC (analog-to-digital converter), which converts the analog signals after photoelectric conversion into digital signals; Offline RX-DSP (offline receive digital signal processing), similar to the offline TX-DSP, but used for data processing on the receiving end Database, used to store, analyze and retrieve large amounts of data.

[0051] In this implementation, three types of soft faults can be considered, including filter bandwidth narrowing (Filtertightening, FT), filter center frequency drift (Filter Shift, FS) and Erbium-Doped Fiber Amplifier Noise Figure Increase (Erbium-Doped Fiber Amplifier Noise Figure Increase, ENFI). The actual system is simulated by introducing random fluctuations of some key parameters, and the parameters of the bandwidth-variable wavelength selector switch (BV-WSS) and EDFA are changed to introduce faults. When a soft fault is introduced, only one related parameter is changed, and other link parameters remain unchanged. The preforward error correction BER corresponding to all the introduced soft faults is lower than the preset threshold. It can be defined as a small sample experiment when the number of training samples for each category does not exceed 50.

[0052] See also Figure 5 , Figure 5 It is a schematic diagram of the model performance experimental results of performing small sample soft fault detection using the first signal feature, the second signal, and the fusion feature of the first signal feature and the second signal in the present invention. Figure 5 The experiment compared the model performance of small sample soft fault detection using prior features (PF), general features (GF), and combined features (CF) in a self-supervised learning scheme based on DCNN. The experimental results are as follows: Figure 5 shown. Figure 5 The index values ​​include F1 score, precision (abbreviated as Prec) and recall. The F1 score comprehensively evaluates the correct detection of soft fault samples and the correct detection of normal samples. The higher the F1 score, the better the model performance.

[0053] See also Figure 6 , Figure 6 It is a schematic diagram of the model performance experimental results of embedding prior knowledge using different self-supervised learning models under small sample conditions in the present invention. Figure 6 Experiment 3 compares the effects of different self-supervised learning models DCNN, MLP, RNN, and models embedded with prior knowledge under small sample conditions. Figure 6 It can be found that when the number of training samples in each category is further reduced, the effect of embedding prior knowledge becomes more significant, which emphasizes the importance of prior knowledge in extremely small sample cases.

[0054] In some optional implementations, the initial fault detection model is fine-tuned in a supervised manner based on the labeled sample signal data and the label data to obtain the fault detection model, including: using a cross entropy loss function to evaluate the difference between the predicted label of the labeled sample signal data and the label data corresponding to the labeled sample signal data, and updating the weights of the initial fault detection model through a back propagation algorithm to minimize the value of the loss function. In addition to the cross entropy loss function, other types of loss functions, such as mean square error, can also be used, depending on the type of model, the goal of the task, and the characteristics of the data.

[0055] In some optional implementations, the characteristic index includes at least one of the following: information theory characteristic index, statistical description characteristic index, morphological characteristic index and ratio index characteristic index; the information theory characteristic index includes information entropy; the statistical description characteristic index includes at least one of the following: mean, variance, standard deviation, kurtosis and skewness; the morphological characteristic index includes at least one of the following: maximum value, minimum value, full width half peak value and symmetry; the ratio index characteristic index includes at least one of the following: waveform index, pulse index and peak index. In addition, the characteristic index may also include some other frequency domain or time series characteristic indexes, which can be selected according to the type of fault to be detected.

[0056] In this implementation, the information entropy p1 can be calculated by the following formula: ,when ; where i and j are natural numbers, x(i) represents the eigenvalue of the i-th sample, x(j) represents the eigenvalue of the j-th sample, N is the number of samples, and p i represents the probability of the i-th sample.

[0057] The mean p2 can be calculated using the following formula: .

[0058] The variance p3 can be calculated by the following formula: .

[0059] The standard deviation p4 can be calculated by the following formula: .

[0060] Kurtosis is a statistic that describes the sharpness of the data distribution. It measures the shape of the data relative to the normal distribution. Kurtosis p5 can be calculated using the following formula: .

[0061] Skewness is a statistic that describes the asymmetry of data distribution. It measures the degree of skewness of the shape of the data distribution relative to the normal distribution. The skewness p6 can be calculated by the following formula: .

[0062] The maximum value p7 can be calculated by the following formula: .

[0063] The minimum value p8 can be calculated by the following formula: .

[0064] Full Width Half Maximum (FWHM) is a term used to describe the width of a function or curve. It indicates the width of a function or curve when it drops to half the height of the peak on both sides of its peak. In signal processing, it can be used to describe the bandwidth of a signal. For a function with symmetrical peaks, FWHM is the distance between the two points where the peak height drops to half the peak height. The full width half maximum value p9 can be calculated using the following formula: ,when .

[0065] Symmetry p 10 It can be calculated by the following formula: the quantized signal at the cutoff frequency f c Asymmetry of power distribution below and above: , where f c Indicates the cutoff frequency.

[0066] Waveform Index p 11 It can be calculated by the following formula: .

[0067] Pulse index p 12 It can be calculated by the following formula: .

[0068] Peak index p 13 It can be calculated by the following formula: .

[0069] The optical network soft fault detection device provided by the present invention is described below. The optical network soft fault detection device described below and the optical network soft fault detection method described above can be referred to each other.

[0070] Figure 7 A schematic diagram of the structure of the optical network soft fault detection device provided in the embodiment of the present application is as follows: Figure 7 As shown, it specifically includes: an acquisition module 701, which is configured to obtain the signal data to be detected of the target optical network; a detection module 702, which is configured to input the signal data to be detected into a pre-trained fault detection model to obtain the fault detection information output by the fault detection model, and the fault detection model is obtained by self-supervised training based on unlabeled sample signal data.

[0071] In this embodiment, the specific processing of the acquisition module 701 and the detection module 702 of the optical network soft fault detection device can be referred to. Figure 2 This corresponds to step 201 and step 202 in the embodiment.

[0072] The optical network soft fault detection device provided by the present invention trains a fault detection model based on a self-supervised training method, completes the training of the fault detection model under the condition of limited training data, and improves the detection efficiency of soft fault detection in the optical network.

[0073] In some optional implementations, the steps of training a fault detection model include: obtaining unlabeled sample signal data; calculating a first signal feature of the unlabeled sample signal data based on a feature index, wherein the feature index is selected based on prior knowledge; extracting a second signal feature from the unlabeled sample signal data based on an encoder; fusing the first signal feature and the second signal feature to obtain a pseudo-label for the unlabeled sample signal data; and training an initial network model based on the unlabeled sample signal data and the pseudo-label to obtain a fault detection model.

[0074] In some optional implementations, extracting the second signal feature from the unlabeled sample signal data based on the encoder includes: using a convolutional autoencoder to automatically learn and extract the second signal feature from the unlabeled sample signal data.

[0075] In some optional implementations, the initial network model is trained based on unlabeled sample signal data and pseudo labels to obtain a fault detection model, including: performing self-supervised training on the initial network model based on the unlabeled sample signal data and the pseudo labels to obtain the initial fault detection model; obtaining labeled sample signal data and label data corresponding to the labeled sample signal data; and performing supervised fine-tuning on the initial fault detection model based on the labeled sample signal data and the label data to obtain the fault detection model.

[0076] In some optional implementations, the initial fault detection model is supervised fine-tuned based on the labeled sample signal data and the label data to obtain the fault detection model, including: using a cross entropy loss function to evaluate the difference between the predicted label of the labeled sample signal data and the label data corresponding to the labeled sample signal data, and updating the weights of the initial fault detection model through a back propagation algorithm to minimize the value of the loss function.

[0077] In some optional implementations, the initial network model is established based on a deep convolutional neural network, which includes a preset number of convolutional layers, a preset number of pooling layers and a fully connected layer. Each convolutional layer is followed by a pooling layer, and finally the prediction result is output through the fully connected layer.

[0078] In some optional implementations, the characteristic indicators include at least one of the following: information theory characteristic indicators, statistical description characteristic indicators, morphological characteristic indicators and ratio index characteristic indicators; the information theory characteristic indicators include information entropy; the statistical description characteristic indicators include at least one of the following: mean, variance, standard deviation, kurtosis and skewness; the morphological characteristic indicators include at least one of the following: maximum value, minimum value, full-width half-peak value and symmetry; the ratio index characteristic indicators include at least one of the following: waveform index, pulse index and peak index.

[0079] Figure 8 An example of a physical structure diagram of an electronic device is shown in FIG. Figure 8 As shown, the electronic device may include: a processor 810, a communications interface 820, a memory 830 and a communication bus 840, wherein the processor 810, the communications interface 820 and the memory 830 communicate with each other through the communication bus 840. The processor 810 may call the logic instructions in the memory 830 to execute the optical network soft fault detection method, which includes: obtaining the signal data to be detected of the target optical network; inputting the signal data to be detected into a pre-trained fault detection model to obtain fault detection information output by the fault detection model, wherein the fault detection model is obtained by self-supervised training based on unlabeled sample signal data.

[0080] In addition, the logic instructions in the above-mentioned memory 830 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art or the part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.

[0081] On the other hand, the present invention also provides a computer program product, which includes a computer program. The computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the optical network soft fault detection method provided by the above methods, and the method includes: obtaining signal data to be detected of the target optical network; inputting the signal data to be detected into a pre-trained fault detection model to obtain fault detection information output by the fault detection model, and the fault detection model is obtained by self-supervised training based on unlabeled sample signal data.

[0082] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, it is implemented to execute the optical network soft fault detection method provided by the above-mentioned methods, the method comprising: obtaining signal data to be detected of the target optical network; inputting the signal data to be detected into a pre-trained fault detection model to obtain fault detection information output by the fault detection model, and the fault detection model is obtained by self-supervised training based on unlabeled sample signal data.

[0083] 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 displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0084] 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 necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product 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 embodiments.

[0085] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention 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 invention.

Claims

1. A method for detecting soft faults in an optical network, characterized in that: include: Acquire the signal data to be detected of the target optical network; The signal data to be detected is input into a pre-trained fault detection model to obtain fault detection information output by the fault detection model, wherein the fault detection model is obtained by self-supervised training based on unlabeled sample signal data.

2. The optical network soft fault detection method according to claim 1, characterized in that: The steps of training the fault detection model include: Obtain unlabeled sample signal data; Calculating a first signal feature of the unlabeled sample signal data according to a feature index, wherein the feature index is selected based on prior knowledge; Extracting a second signal feature from the unlabeled sample signal data based on an encoder; The first signal feature and the second signal feature are combined to obtain a pseudo label for the unlabeled sample signal data; The initial network model is trained based on the unlabeled sample signal data and the pseudo labels to obtain the fault detection model.

3. The optical network soft fault detection method according to claim 2, characterized in that: The extracting a second signal feature from the unlabeled sample signal data based on an encoder includes: A convolutional autoencoder is used to automatically learn and extract second signal features from the unlabeled sample signal data.

4. The optical network soft fault detection method according to claim 2, characterized in that: The training of the initial network model based on the unlabeled sample signal data and the pseudo label to obtain the fault detection model includes: Performing self-supervisory training on the initial network model based on the unlabeled sample signal data and the pseudo labels to obtain an initial fault detection model; Acquire labeled sample signal data and label data corresponding to the labeled sample signal data; The initial fault detection model is fine-tuned in a supervised manner based on the labeled sample signal data and the label data to obtain the fault detection model.

5. The optical network soft fault detection method according to claim 4, characterized in that: The supervised fine-tuning of the initial fault detection model based on the labeled sample signal data and the label data to obtain the fault detection model includes: A cross entropy loss function is used to evaluate the difference between the predicted label of the labeled sample signal data and the label data corresponding to the labeled sample signal data, and a back propagation algorithm is used to update the weight of the initial fault detection model to minimize the value of the loss function.

6. The optical network soft fault detection method according to claim 1, characterized in that: The characteristic indicators include at least one of the following: information theory characteristic indicators, statistical description characteristic indicators, morphological characteristic indicators and ratio index characteristic indicators; the information theory characteristic indicators include information entropy; the statistical description characteristic indicators include at least one of the following: mean, variance, standard deviation, kurtosis and skewness; the morphological characteristic indicators include at least one of the following: maximum value, minimum value, full width half peak value and symmetry; the ratio index characteristic indicators include at least one of the following: waveform index, pulse index and peak index.

7. An optical network soft fault detection device, characterized in that: include: An acquisition module is configured to acquire the signal data to be detected of the target optical network; The detection module is configured to input the signal data to be detected into a pre-trained fault detection model to obtain fault detection information output by the fault detection model, wherein the fault detection model is obtained by self-supervised training based on unlabeled sample signal data.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the optical network soft fault detection method according to any one of claims 1 to 6 is implemented.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the optical network soft failure detection method according to any one of claims 1 to 6 is implemented.

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

Images