Optical fiber vibration monitoring and evaluating method and device

Abstract

The invention relates to an optical fiber vibration monitoring and evaluating method and device, and the method comprises the steps: obtaining a time domain signal sample, and converting the time domain signal sample into a signal feature sample; inputting the signal feature sample into a trained Stacking model to obtain an event category prediction result; wherein the Stacking model comprises a base learner stage and a model fusion stage, the base learner stage comprises a plurality of base learners, and the model fusion stage comprises an integrated learner; the trained Stacking model is obtained by training a training set, the training set is composed of a signal feature sample converted from a time domain signal sample and a vibration event label, the signal feature sample is input into a base learner stage, a plurality of base learners give event category prediction vectors for the signal feature sample, and the event category prediction vectors are input into the base learners; and taking the output of the plurality of base learners and the vibration event labels of the corresponding samples as the input of an integrated learner in a model fusion stage, and training the integrated learner.

Description

Technical Field

[0001] The present application relates to the technical field of optical fiber fault detection, and in particular to an optical fiber vibration monitoring and evaluation method and device. Background Art

[0002] As the main medium for data transmission, optical fiber is increasingly used with the continuous development of the information age. The popularity and extended service life of optical fiber make it essential to monitor and evaluate its health status. Ensuring the good operation of optical fiber is not only the key to maintaining data transmission, but also a necessary measure to prevent communication failures and data loss. Some human damage or bad construction behavior are very likely to affect the normal operation of optical fiber. Therefore, it is extremely necessary to promptly detect third-party intrusion incidents in the environment where the optical fiber is located.

[0003] In the existing technology, the detection of third-party intrusion events mainly relies on vibration source detection. One of the means is to use the sensor data of the phase-sensitive optical time domain reflectometer (Φ-OTDR) and use different machine learning and deep learning methods to identify patterns in the sensor data, so as to find dangerous vibration events, identify them as third-party intrusion events and issue early warnings.

[0004] Among these machine learning and deep learning methods, they can be mainly divided into supervised learning represented by SVM, random forest, CNN, and LSTM, and unsupervised learning represented by KNN and DBSCAN. The former can classify vibration events with different patterns. Each model has its advantages in specific types of data or specific problems, but a single model may be too complex and tend to overfit the data; while the latter, although the model is simple and robust, the recognition accuracy of samples at the edge of a cluster is often low.

[0005] There are also methods of fusing multiple models in the prior art, such as CN116992330A, an unbalanced data fault diagnosis and identification method and device based on the Stacking integration strategy, which includes multiple classifiers, and individual classifiers include support vector machine SVM, gradient boosting decision tree GBDT, extreme gradient boosting XGBoost, lightweight gradient boosting machine LightGBM, artificial neural network ANN, logistic regression LRA, K nearest neighbor algorithm KNN. For example, CN115410643A is a prediction method, system, device and storage medium for enhancers based on stacking ensemble learning. The K nearest neighbor (KNN) algorithm is used to construct 5 base classifiers with different parameters as the base model of the stacking ensemble learning model. Different base classifiers are independent of each other and are calculated in parallel during training. The present invention proposes a method and device for monitoring and evaluating optical fiber vibration that is different from the above method, which can adjust parameters for different scenarios to meet the actual use requirements of different scenarios. Summary of the invention

[0006] The present invention aims to provide a method and device for monitoring and evaluating optical fiber vibration to solve the deficiencies in the prior art. The technical problem to be solved by the present invention is achieved through the following technical solutions.

[0007] An optical fiber vibration monitoring and evaluation method, comprising:

[0008] Acquire time domain signal samples, and convert the time domain signal samples into signal feature samples;

[0009] Input the signal feature samples into the trained Stacking model to obtain the classification label;

[0010] The Stacking model includes a base learner stage and a model fusion stage, wherein the base learner stage includes multiple base learners and the model fusion stage includes an integrated learner;

[0011] The trained Stacking model is obtained by training the training set, specifically:

[0012] The training set consists of signal feature samples and vibration event labels converted from time domain signal samples. The signal feature samples are input into the base learner stage. Multiple base learners give event category prediction vectors for the signal feature samples respectively. The outputs of multiple base learners and the vibration event labels of the corresponding samples are used as the input of the integrated learner in the model fusion stage to train the integrated learner.

[0013] The time domain signal samples are sensing data based on Φ-OTDR technology.

[0014] The base learner stage includes five base learners, namely KNN, LightGBM, XGBoost, AdaBoost and bagging classifier, and the ensemble learner is XGBoost classifier.

[0015] The signal feature sample is a vector composed of indicator features, including four indicators, namely basic statistical features, volatility and shape features, energy and proportion features, and dynamic range features. The basic statistical features are used to indicate the central trend and fluctuation range of the signal, the volatility and shape features are used to indicate the shape and volatility of the signal distribution, the energy and proportion features are used to indicate the energy and peak value of the signal, and the dynamic range features are used to indicate the difference between the maximum amplitude and the minimum amplitude of the signal and the dynamic range of the signal.

[0016] The signal feature sample is The corresponding classification label is Where k is the number of samples of vibration events, m is the number of features of each sample, and each base learner will obtain k prediction vectors for these k samples, which are combined and expressed as r is the number of vibration event categories. Each sample will get a probability 1xr belonging to each category, and the sum of the vector elements of each 1xr is set to 1. Then all the prediction results of the base learners are stacked to form n is the number of base learners.

[0017] Multiple base learners are stacked to form Input into the integrated learner for training and output classification labels.

[0018] The ensemble learner compares the predicted classification results of a sample by multiple base learners with the actual classification results of the sample to determine the prediction performance of different base learners in different situations, thereby adjusting parameters for different scenarios.

[0019] The ensemble learner judges the performance of the model based on the accuracy, precision, recall, F1 score and AUC indicators, where accuracy is the ratio of correct predictions of the overall samples, precision is the proportion of true positive examples among the samples predicted as positive examples, recall is the proportion of predicted positive examples among the true positive examples, F1 score is the harmonic mean of precision and recall, and AUC is the area under the ROC curve.

[0020] An optical fiber vibration monitoring and evaluation device, comprising:

[0021] A sample acquisition module acquires time domain signal samples and converts the time domain signal samples into signal feature samples;

[0022] Stacking model, receiving signal feature samples, inputting the signal feature samples into the Stacking model to obtain classification labels;

[0023] The Stacking model includes a base learner stage and a model fusion stage, wherein the base learner stage includes multiple base learners and the model fusion stage includes an integrated learner;

[0024] The Stacking model is obtained by training the training set, specifically:

[0025] The training set consists of signal feature samples and vibration event labels converted from time domain signal samples. The signal feature samples are input into the base learner stage. Multiple base learners give event category prediction vectors for the signal feature samples respectively. The outputs of multiple base learners and the vibration event labels of the corresponding samples are used as the input of the integrated learner in the model fusion stage to train the integrated learner.

[0026] The base learner stage of the Stacking model includes five base learners, namely KNN, LightGBM, XGBoost, AdaBoost and bagging classifier, and the ensemble learner is XGBoost classifier.

[0027] The optical fiber vibration monitoring and evaluation method of the present invention has the following beneficial effects:

[0028] 1. By combining the predictions of multiple different machine learning models, it is not only possible to process and analyze large amounts of vibration data, but also to effectively improve the accuracy and efficiency of vibration event identification by integrating the prediction results of multiple models.

[0029] 2. The ensemble learner compares the predicted classification results of the base learners with the actual classification results of the samples to determine the prediction performance of different base learners under different circumstances, thereby adjusting the parameters for different scenarios.

[0030] 3. The integrated learner judges the performance of the model through accuracy, precision, recall, F1 score and AUC indicators, and sets corresponding weights according to the requirements of different scenarios to meet the actual usage needs of different scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 Schematic diagram of the structure of the optical fiber vibration monitoring and evaluation device of the present invention DETAILED DESCRIPTION

[0032] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments.

[0033] Example

[0034] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.

[0035] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention claimed for protection, but merely represents selected embodiments of the present invention. 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.

[0036] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, further definition and explanation thereof is not required in subsequent drawings.

[0037] In the description of the present invention, it should be noted that the terms "upper", "lower", "inside", "outside", etc. indicate directions or positional relationships based on the directions or positional relationships shown in the accompanying drawings, or are directions or positional relationships in which the product of the invention is usually placed when in use. They are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore should not be understood as a limitation on the present invention.

[0038] In the description of the present invention, it is also necessary to explain that, unless otherwise clearly specified and limited, the terms "setting", "installation" and "connection" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be an indirect connection through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0039] The embodiments of the present invention are described in detail below in conjunction with the accompanying drawings.

[0040] An optical fiber vibration monitoring and evaluation method, comprising:

[0041] Get time domain signal samples,

[0042] The time domain signal sample is the sensing data based on Φ-OTDR technology. The working principle of Φ-OTDR technology is that when external disturbance acts on the sensing optical fiber, it will change the refractive index of the sensing optical fiber, causing the Rayleigh scattered light to produce phase modulation. Distributed sensing can be performed by demodulating the intensity or phase information of the backward Rayleigh scattered light pulse signal in the optical fiber. After the coherent pulse light is injected into the sensing optical fiber through the circulator, the generated backward Rayleigh scattered light returns to the front end of the optical fiber, is received by the photoelectric detector through the circulator, and is demodulated by the demodulation unit to restore the external strain change information. The intensity demodulation type OTDR technology directly uses the photoelectric detector for intensity detection, which is used to locate the external strain; the phase demodulation type Φ-OTDR / DAS technology uses interferometer demodulation or coherent demodulation method. Since the external strain change is proportional to the phase of the interference signal, this method can quantitatively restore the size, frequency and phase of the external strain change.

[0043] Convert time domain signal samples into signal feature samples;

[0044] The signal feature sample is a vector composed of indicator features. Each time domain signal sample is converted into a vector composed of indicator features, which reflect the main properties of this sample. This set of indicator features mainly includes four aspects: first, basic statistical features, which can capture the central trend and fluctuation range of the signal, and are the basis for identifying different types of vibration events; second, volatility and shape features, which reveal the shape and volatility of the signal distribution, and can reflect the discreteness, sharpness, asymmetry and other information of the signal; third, energy and ratio features, which describe the energy characteristics and peak characteristics of the signal, and are particularly effective for identifying events with concentrated energy; fourth, dynamic range features, which provide the difference between the maximum amplitude and the minimum amplitude of the signal, and help the model understand the dynamic range of the signal.

[0045] Input the signal feature samples into the trained Stacking model to obtain the classification label;

[0046] Combination Figure 1 As shown, the Stacking model includes a base learner stage and a model fusion stage, wherein the base learner stage includes multiple base learners, and the model fusion stage includes an integrated learner;

[0047] The trained Stacking model is obtained by training the training set, specifically:

[0048] The training set consists of signal feature samples and vibration event labels converted from time domain signal samples. The signal feature samples are input into the base learner stage. Multiple base learners give event category prediction vectors for the signal feature samples respectively. The outputs of multiple base learners and the vibration event labels of the corresponding samples are used as the input of the integrated learner in the model fusion stage to train the integrated learner.

[0049] This model is based on more than ten machine learning models with different principles and complexities. Five models, including K-nearest neighbor, LightGBM, XGBoost, AdaBoost and bagging classifier, are selected as the models in the base learner stage, and XGBoost classifier is selected as the model in the model fusion stage.

[0050] Assume that the signal feature sample is The corresponding classification label is k is the number of samples of vibration events, m is the number of features of each sample, and x is input into five base learners to predict their respective classification labels. Specifically, each classifier will obtain k prediction vectors for these k samples, which are combined and expressed as r is the number of categories of all vibration events stored in the database. Each sample will get a probability of belonging to each category 1xr, and the sum of each vector element of 1xr is 1.

[0051] For example, at this stage, the five base learners will be given the vibration signal features and output their own predictions for the probability of each feature belonging to a category. For example, assuming there are 4 categories in total, the prediction result of a base learner for a certain feature is a 1x4 vector, which can be [0.05, 0.06, 0.04, 0.85].

[0052] Finally, all the classifier prediction results are stacked to form Where k is the number of samples, r is the number of vibration event categories, and n represents a total of n base learners involved in the prediction. Input into the integrated learner for training and output classification labels.

[0053] The ensemble learner compares the predicted classification results of a sample by multiple base learners with the actual classification results of the sample to determine the prediction performance of different base learners in different situations, thereby adjusting parameters for different scenarios.

[0054] In actual application scenarios, for the protection of critical infrastructure, such as large stations in railway networks, core nodes in optical fiber communication networks, and high-level substations in power supply and transmission networks, these places are connected to other nodes through optical cable networks. For these high-risk scenarios, it is crucial to accurately identify all high-risk behaviors. The monitoring requirements are not only to identify all possible threats in a timely and accurate manner, but also to issue early warnings in the early stages of intrusions to prevent any potential damage. In other places, such as the daily maintenance of optical fiber communication networks, frequent false alarms will consume a lot of human resources for unnecessary investigations, affecting operation and maintenance efficiency, so the recognition accuracy should be improved as much as possible and false alarms should be reduced.

[0055] The ensemble learner judges the performance of the model based on the accuracy, precision, recall, F1 score and AUC indicators, where accuracy is the ratio of correct predictions of the overall samples, precision is the proportion of true positive examples among the samples predicted as positive examples, recall is the proportion of predicted positive examples among the true positive examples, F1 score is the harmonic mean of precision and recall, and AUC is the area under the ROC curve.

[0056] Based on the trained base learners, the next step is to build an ensemble learner in the model fusion stage. The goal of this stage is to let the ensemble learner learn the prediction rules of each base learner in different situations, use the prediction diversity of the base learners, grasp the advantages and disadvantages of each base learner in different given situations, and learn from the prediction and classification deviations of the underlying model, so as to obtain higher prediction accuracy.

[0057] Based on the base learner stage, the XGBoost model in the model fusion stage will accept the results of each of the five base learners. In this process, the XGBoost model in the model fusion stage plays the role of a meta-learner. In the base learner stage, after each base learner makes a prediction on the training data, their output (such as predicted probability or category label) will be used as the input features of the XGBoost model. After the base learner stage is over, the prediction results output from the base learners are It will be input into the final XGBoost-based stacked classifier for further training, and finally output the predicted classification results.

[0058] To illustrate with a specific example, if the prediction results of the five base learners for a certain sample are [0.05, 0.06, 0.04, 0.85], [0.15, 0.66, 0.04, 0.15], ... (five in total). These results reflect the different understandings and judgments of the data point by different models. The XGBoost model will learn the patterns behind these prediction results. In the prediction process, it is possible that the five base learners have different prediction results for a sample. For example, in the example, the first base learner believes that it should belong to the fourth category (0.85), and the second base learner believes that it should belong to the second category (0.66). At this time, the second-order XGBoost model can compare the prediction results of the five base learners for the sample with the actual classification results of the sample during the learning process, so as to judge which base learners' predictions are credible and which may be misjudgments in different situations.

[0059] Based on the signal sample characteristics, five model evaluation indicators are used to measure the performance of the model, including:

[0060] Accuracy is the ratio of correct predictions of the overall sample, that is, the proportion of all vibration events that are correctly identified as third-party intrusions or non-intrusions:

[0061]

[0062] Precision is the proportion of true positive examples among samples predicted to be positive examples. For fiber optic vibration detection, high precision means that when an alarm is triggered, there is a high probability that it is a real threat, which is critical to avoid wasting resources on false alarms:

[0063]

[0064] Among them, N represents the number of vibration types, Precisioni represents the accuracy of category i, true positive examples (TPi) represent the number of samples correctly predicted as category i, and false positive examples (FPi) represent the number of samples incorrectly predicted as category i (actually belonging to other categories but predicted as i).

[0065] Recall is the ratio of positive examples predicted in real positive samples, reflecting the coverage of the model's correct prediction of dangerous events. In fiber security monitoring, a high recall rate ensures that almost all real intrusion events can be detected, thereby preventing any potential data or service interruptions:

[0066]

[0067] Where N represents the number of vibration types, Recalli represents the recall rate of category i, true positive examples (TPi) represent the number of samples correctly predicted as category i, and false negative examples (FNi) represent the number of samples that are actually category i but are incorrectly predicted as other categories.

[0068] The F1 score (F1_score) is the harmonic mean of precision and recall. A high F1 score indicates that the model does not miss real intrusion events while maintaining a low false alarm rate:

[0069]

[0070] AUC (Area Under Curve) is the area under the ROC curve (Receiver Operating Characteristic Curve), which comprehensively reflects the classification performance of the model. For various types of fiber vibration, it is necessary to calculate the AUC corresponding to each type of ROC curve and take the average. In fiber vibration detection, a higher AUC value indicates that the model has excellent classification performance and can effectively identify real threats.

[0071]

[0072] In this stacked model, XGBoost’s task is to synthesize this information, extract the factors that are most conducive to the final decision, and adjust and optimize the final prediction results through its powerful classification ability.

[0073] Define indicator models for different scenarios

[0074] (1) High-risk behavior sensitive scenarios

[0075] In such scenarios, the model must be extremely sensitive to high-risk behaviors, even at the cost of misidentifying some safe behaviors as high-risk behaviors. Although such misidentification will increase some unnecessary alarms, the waste of resources caused by false alarms is acceptable compared to missing real high-risk events. The core of this strategy is to ensure that high-risk events are not ignored, thereby ensuring the security and stability of the system.

[0076] The application of the Stacking model in this scenario can take advantage of its combination of multiple base models to improve the ability to identify abnormal vibration patterns through multi-level and multi-angle analysis. By merging the prediction results of different models, the Stacking model can maintain high sensitivity while minimizing the false alarm rate, thereby providing reliable protection in high-risk scenarios.

[0077] In high-risk scenarios, the key indicators are recall and F1 score. This is because in this scenario, the key is to capture all potential high-risk events as much as possible, even if this will bring some false positives.

[0078] Recall: Recall is the most important indicator in high-risk scenarios. It reflects the model's coverage of true positive samples (i.e., high-risk events). In high-risk scenarios, it is unacceptable to miss any high-risk event, so the model needs to have a high recall rate to ensure that almost all high-risk events can be detected.

[0079] F1 Score: The F1 score is the harmonic mean of precision and recall. In high-risk scenarios, although recall is crucial, precision must also be considered to avoid too many false positives. Therefore, the F1 score, as a comprehensive indicator, can balance the importance of the two and ensure that the model maintains a high precision while maintaining a high recall.

[0080] (2) Ordinary scenes

[0081] In this scenario, the fiber optic vibration monitoring model needs to use precise feature extraction and classification algorithms to ensure that only truly abnormal events can trigger alarms. The Stacking model can balance accuracy and sensitivity in common scenarios and optimize the performance of the alarm system by integrating the prediction results of multiple base models. For example, by using analysis based on multi-dimensional features such as energy features and dynamic range features, the model's ability to distinguish between normal and abnormal events can be improved, reducing the occurrence of false alarms.

[0082] In common scenarios, the model needs to focus on precision and accuracy to reduce false positives and improve overall recognition accuracy.

[0083] Precision: In common scenarios, false alarms consume a lot of human resources for unnecessary investigation. Therefore, precision is a key indicator, which reflects the proportion of true positive samples among samples predicted as positive samples. High precision means that most alarms are valid, which can reduce resource waste.

[0084] Accuracy: Accuracy is the ratio of correct predictions for the overall samples. In common scenarios, the accuracy of overall monitoring is equally important. High accuracy means that the model can correctly identify vibration events in most cases, reducing false positives and false negatives.

[0085] Set indicator weights for different scenarios

[0086] In order to optimize model performance in different scenarios, the weights of each indicator can be preset according to the importance of the scenario and the model training results.

[0087] (1) High-risk behavior sensitive scenarios

[0088] In high-risk scenarios, the weight of recall and F1 score is increased. This weight setting ensures that the model can capture high-risk events to the greatest extent in high-risk scenarios and minimize false positives while capturing them.

[0089] For example:

[0090] Recall: 50%

[0091] F1 Score: 30%

[0092] Accuracy: 10%

[0093] Precision: 5%

[0094] AUC: 5%

[0095] (2) Ordinary scenes

[0096] In common scenarios, the weights of precision and accuracy are increased. This weight setting ensures that the model can reduce false positives in common scenarios and improve the effectiveness and accuracy of overall monitoring.

[0097] For example:

[0098] Precision: 40%

[0099] Accuracy: 30%

[0100] F1 Score: 15%

[0101] Recall: 10%

[0102] AUC: 5%

[0103] Finally, by adjusting the weighted scores of the five indicators, the performance of the model for different application scenarios is obtained. The advantage of this method is that it can dynamically adjust the weight of the base learner in the final prediction according to its performance, which not only improves the accuracy of the model, but also enhances the model's ability to adapt to the inherent complexity of the data.

[0104] The present invention also relates to an optical fiber vibration monitoring and evaluation device, comprising:

[0105] A sample acquisition module acquires time domain signal samples and converts the time domain signal samples into signal feature samples;

[0106] The Stacking model receives signal feature samples and inputs the signal feature samples into the Stacking model to obtain event category prediction results;

[0107] The Stacking model includes a base learner stage and a model fusion stage, wherein the base learner stage includes multiple base learners and the model fusion stage includes an integrated learner;

[0108] The Stacking model is obtained by training the training set, specifically:

[0109] The training set consists of signal feature samples and vibration event labels converted from time domain signal samples. The signal feature samples are input into the base learner stage. Multiple base learners give event category prediction vectors for the signal feature samples respectively. The outputs of multiple base learners and the vibration event labels of the corresponding samples are used as the input of the integrated learner in the model fusion stage to train the integrated learner.

[0110] The base learner stage of the Stacking model includes five base learners, namely KNN, LightGBM, XGBoost, AdaBoost and bagging classifier, and the ensemble learner is XGBoost classifier.

[0111] During the application process, the Stacking model analyzes the similarity between the vibration events predicted by each base model and the known third-party intrusion patterns, and the advanced learner integrates this information to automatically identify and classify the nature of the current vibration event. This model is not only suitable for static data analysis, but also for real-time monitoring systems. It can timely detect and warn possible third-party intrusion events, providing an efficient and reliable technical means for the security protection of optical fiber communication networks. By implementing the present invention, the security monitoring level of optical fiber communication networks can be significantly improved, communication failures caused by third-party intrusions can be effectively prevented, and the stability and reliability of data transmission can be ensured.

[0112] It should be noted that the above detailed descriptions are exemplary and are intended to provide further explanation of the present application. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the art to which the present application belongs.

[0113] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments described in the present application. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should also be understood that when the terms "comprise" and / or "include" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0114] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the terms used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein.

[0115] In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units that are not explicitly listed or inherent to these processes, methods, products, or apparatuses.

[0116] For ease of description, spatially relative terms, such as "above", "above", "on the upper surface of", "above", etc., may be used herein to describe the spatial positional relationship between a device or feature and other devices or features as shown in the figure. It should be understood that spatially relative terms are intended to include different orientations of the device in use or operation in addition to the orientation described in the figure. For example, if the device in the accompanying drawings is inverted, the device described as "above other devices or structures" or "above other devices or structures" will be positioned as "below other devices or structures" or "below other devices or structures". Thus, the exemplary term "above" may include both "above" and "below". The device may also be positioned in other different ways, such as rotated 90 degrees or in other orientations, and the spatially relative descriptions used herein are interpreted accordingly.

[0117] In the above detailed description, reference is made to the accompanying drawings, which form a part of this document. In the accompanying drawings, similar symbols typically identify similar components unless the context indicates otherwise. The illustrated embodiments described in the detailed description, drawings, and claims are not meant to be limiting. Other embodiments may be used, and other changes may be made, without departing from the spirit or scope of the subject matter presented herein.

[0118] The above description is only the preferred embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A method for monitoring and evaluating optical fiber vibration, characterized in that: include: Acquire time domain signal samples, and convert the time domain signal samples into signal feature samples; Input the signal feature samples into the trained Stacking model to obtain the classification label; The Stacking model includes a base learner stage and a model fusion stage, wherein the base learner stage includes multiple base learners and the model fusion stage includes an integrated learner; The trained Stacking model is obtained by training the training set, specifically: The training set consists of signal feature samples and vibration event labels converted from time domain signal samples. The signal feature samples are input into the base learner stage. Multiple base learners give event category prediction vectors for the signal feature samples respectively. The outputs of multiple base learners and the vibration event labels of the corresponding samples are used as the input of the integrated learner in the model fusion stage to train the integrated learner.

2. The optical fiber vibration monitoring and evaluation method according to claim 1, characterized in that: The time domain signal samples are sensing data based on Φ-OTDR technology.

3. The optical fiber vibration monitoring and evaluation method according to claim 2, characterized in that: The base learner stage includes five base learners, namely KNN, LightGBM, XGBoost, AdaBoost and bagging classifier, and the ensemble learner is XGBoost classifier.

4. The optical fiber vibration monitoring and evaluation method according to claim 1, characterized in that: The signal feature sample is a vector composed of indicator features, including four indicators, namely basic statistical features, fluctuation and shape features, energy and proportion features, and dynamic range features. The basic statistical features are used to indicate the central trend and fluctuation range of the signal, the fluctuation and shape features are used to indicate the shape and volatility of the signal distribution, the energy and proportion features are used to indicate the energy and peak value of the signal, and the dynamic range features are used to indicate the difference between the maximum amplitude and the minimum amplitude of the signal and the dynamic range of the signal.

5. The optical fiber vibration monitoring and evaluation method according to claim 1, characterized in that: The signal feature sample is The corresponding classification label is Where k is the number of samples of vibration events, m is the number of features of each sample, and each base learner will obtain k prediction vectors for these k samples, which are combined and expressed as r is the number of vibration event categories. Each sample will get a probability 1xr belonging to each category, and the sum of the vector elements of each 1xr is set to 1. Then all the prediction results of the base learners are stacked to form n is the number of base learners.

6. The optical fiber vibration monitoring and evaluation method according to claim 5, characterized in that: Multiple base learners are stacked to form Input into the integrated learner for training and output classification labels 7. The optical fiber vibration monitoring and evaluation method according to claim 1, characterized in that: The ensemble learner compares the classification labels of samples by multiple base learners with the vibration event labels of the samples to determine the prediction performance of different base learners in different situations, thereby adjusting parameters for different scenarios.

8. The optical fiber vibration monitoring and evaluation method according to claim 7, characterized in that: The ensemble learner judges the performance of the model based on the accuracy, precision, recall, F1 score and AUC indicators, where accuracy is the ratio of correct predictions of the overall samples, precision is the proportion of true positive examples among the samples predicted as positive examples, recall is the proportion of predicted positive examples among the true positive examples, F1 score is the harmonic mean of precision and recall, and AUC is the area under the ROC curve.

9. An optical fiber vibration monitoring and evaluation device, characterized in that: include: A sample acquisition module acquires time domain signal samples and converts the time domain signal samples into signal feature samples; Stacking model, receiving signal feature samples, inputting the signal feature samples into the Stacking model to obtain classification labels; The Stacking model includes a base learner stage and a model fusion stage, wherein the base learner stage includes multiple base learners and the model fusion stage includes an integrated learner; The Stacking model is obtained by training the training set, specifically: The training set consists of signal feature samples and vibration event labels converted from time domain signal samples. The signal feature samples are input into the base learner stage. Multiple base learners give event category prediction vectors for the signal feature samples respectively. The outputs of multiple base learners and the vibration event labels of the corresponding samples are used as the input of the integrated learner in the model fusion stage to train the integrated learner.

10. The optical fiber vibration monitoring and evaluation device according to claim 9, characterized in that: The base learner stage of the Stacking model includes five base learners, namely KNN, LightGBM, XGBoost, AdaBoost and bagging classifier, and the ensemble learner is XGBoost classifier.

Images