Optical fiber fault prediction and early warning method based on machine learning

Abstract

The invention discloses an optical fiber fault prediction and early warning method based on machine learning. Comprising the steps of obtaining multi-dimensional optical feature data of an optical fiber; labeling the fault area to obtain labeled feature data; preprocessing the marked feature data, and constructing a standard optical fiber fault feature data set; inputting the time sequence optical feature data into a neural network model to obtain marked feature data of each time period and future, and correcting the marked feature data through a standard data set; and inputting real-time data of a to-be-detected optical fiber into the model, and outputting fault data and prediction data. According to the method, multi-dimensional features are processed through machine learning, accurate prediction and early warning of optical fiber faults are achieved, and timeliness and accuracy of fault detection are improved.

Description

Technical Field

[0001] The present invention relates to the technical field, and in particular to a method for predicting and warning optical fiber faults based on machine learning. Background Art

[0002] In fiber-optic communication systems, optical fiber serves as the core carrier of information transmission, and the stability of its operating state directly determines the quality of communication. However, over long-term use, optical fiber is susceptible to environmental factors, physical damage, and aging, leading to various faults such as fiber breakage and abnormal attenuation. These faults can lead to communication interruptions or degradation of signal quality, causing significant inconvenience to production and daily life.

[0003] Currently, fiber optic fault detection primarily relies on manual inspections and traditional monitoring methods. Manual inspections are not only time-consuming and labor-intensive, but also difficult to monitor in real time. Problems are often discovered only after they occur, making early prediction and early warning impossible. Traditional monitoring methods typically only monitor single characteristic parameters, such as optical power and attenuation coefficient, and fail to comprehensively consider the multi-dimensional characteristics of optical fibers. This results in low fault identification accuracy and an inability to effectively predict fault development trends.

[0004] With the development of machine learning technology, it has shown great potential in the field of fault prediction and diagnosis. Applying machine learning technology to optical fiber fault prediction and early warning can comprehensively process the multi-dimensional characteristic data of optical fibers, improving the accuracy of fault identification and the timeliness of prediction. Therefore, developing a machine learning-based optical fiber fault prediction and early warning method is of great practical significance. Summary of the Invention

[0005] To address the above problems, the present invention provides a fiber optic fault prediction and early warning method based on machine learning, which realizes early prediction and accurate early warning of faults through the fusion of temporal and spatial features and dynamic model correction.

[0006] The purpose of the present invention is achieved by adopting the following technical solutions: A fiber optic fault prediction and early warning method based on machine learning includes the following steps: S1. Acquire multi-dimensional optical characteristic data of the optical fiber; S2. marking the fault area of ​​the multi-dimensional optical feature data to obtain marked feature data; S3, preprocessing the annotated feature data to construct a standard optical fiber fault feature data set; S4, inputting the time-series optical feature data in the multidimensional optical feature data into a neural network model, obtaining the annotated feature data of each time period in the time-series optical feature and the predicted annotated feature data of the future time period, and correcting them with a standard optical fiber fault feature data set; S5. Input the multi-dimensional optical characteristic data sampled in real time from the optical fiber to be tested into the neural network model of S4, and output the optical fiber fault data and prediction data.

[0007] As a preferred embodiment, the multi-dimensional optical feature data includes at least deformation features and transmission features; The deformation characteristics are determined according to the surface curvature of the optical fiber, comprising the following steps: Obtaining the surface curvature of each point on the surface and inside of the optical fiber, and determining whether the surface curvature at each point is within a preset threshold range; If it is not within the preset threshold range, obtaining the surface curvature of the adjacent points of the current point until obtaining a point whose surface curvature is within the preset threshold range; Acquire multiple abnormal deformation parts formed by the point set to obtain the deformation feature; The transmission characteristics are determined according to the transmission parameters of the optical fiber, and include the following steps: Acquire an optical signal transmitted from the deformable feature portion to another deformable feature portion or a non-deformable feature portion, and determine a current optical attenuation coefficient; Obtaining a light attenuation coefficient sequence, wherein each unit of the sequence is set to the light attenuation coefficient and transmission distance from the current deformation feature portion to one of the deformation feature portions or the non-deformation feature portion; Obtaining a light attenuation coefficient set, wherein the light attenuation coefficient set is configured to include a sequence of light attenuation coefficients of each of the deformation feature portions; The transmission characteristics are obtained by comparing each unit of the light attenuation coefficient set with the light attenuation coefficient threshold interval preset in the transmission parameters.

[0008] As a preferred method, marking the fault area in step S2 includes the following steps: When the surface curvature deviation of the abnormally deformed part exceeds the preset threshold and the corresponding light attenuation coefficient exceeds the threshold range but does not fluctuate significantly, it is marked as the first fault type; When the difference in the light attenuation coefficients of adjacent units in the light attenuation coefficient sequence exceeds the preset fluctuation threshold, and the surface curvature deviation of the abnormal deformation part is within the allowable range, it is marked as the second fault type; When the light attenuation coefficients corresponding to the abnormal deformation parts exceeding the preset ratio all exceed the threshold range, and the light attenuation coefficients corresponding to the transmission distance are generally high, it is marked as the third fault type.

[0009] As a preferred embodiment, the pretreatment includes: Eliminate coordinate points in deformation characteristics that exceed the physical size range of the optical fiber, and abnormal optical attenuation coefficients outside the distortion range in transmission characteristics; The surface curvature value in the deformation feature is normalized to the interval [0, 1] based on the curvature threshold of the standard optical fiber, and the optical attenuation coefficient in the transmission feature is normalized based on the theoretical optimal attenuation rate; A mapping relationship between the coordinates of the abnormal deformation part and the corresponding light attenuation coefficient sequence is established to form an associated feature vector.

[0010] As a preferred embodiment, the neural network model in step S4 includes: The time series feature extraction layer uses a long short-term memory network to extract time-related features from the time series optical feature data, including the time series data of the optical attenuation coefficient sequence changes and the position expansion time series data of the deformation abnormal part within a continuous time period; The spatial feature extraction layer uses a convolutional neural network to extract the spatial distribution features of the three-dimensional coordinates of the abnormal deformation part in the deformation feature; The fusion prediction layer fuses the outputs of the temporal feature extraction layer and the spatial feature extraction layer, and outputs the labeled feature data of each time period and the predicted labeled feature data of the future time period.

[0011] As a preferred embodiment, the correction using the standard optical fiber fault feature data set in step S4 includes the following steps: Compare the predicted and marked feature data with the features of the same type of fault in the standard optical fiber fault feature data set, and calculate the feature deviation value; If the deviation value exceeds the preset correction threshold, the weight parameters of the neural network model are adjusted through the back propagation algorithm, and the correction is repeated until the deviation value is within the preset allowable range.

[0012] As an optimal method, when outputting prediction data, the impact of the expansion of the deformation area on the transmission coefficient is simulated, the three-dimensional coordinate expansion range of the abnormal deformation part and the corresponding transmission coefficient change value within the future preset time period are predicted, and the prediction results of the fault development trend are output.

[0013] The beneficial effects of the present invention are: The present invention can comprehensively reflect the state of the optical fiber by acquiring multi-dimensional optical characteristic data of the optical fiber, including deformation characteristics and transmission characteristics, and provide a rich basis for fault prediction and early warning.

[0014] Accurately marking the fault area and classifying different fault types can help improve the accuracy of fault recognition by the neural network model.

[0015] Preprocessing the labeled feature data, removing abnormal data and normalizing it, and establishing associated feature vectors can improve the quality and consistency of the data and lay a good foundation for model training and prediction.

[0016] The neural network model including the time series feature extraction layer, the spatial feature extraction layer and the fusion prediction layer can effectively fuse the time series and spatial features and improve the fault prediction ability.

[0017] The prediction results were corrected using a standard optical fiber fault feature dataset, and the model parameters were continuously optimized to further improve the accuracy of the prediction.

[0018] It can output the prediction results of fault development trends, realize the early prediction and early warning of optical fiber faults, and facilitate staff to take timely measures to reduce the losses caused by faults. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The present invention is further described with reference to the accompanying drawings. However, the embodiments in the accompanying drawings do not constitute any limitation to the present invention. A person skilled in the art can obtain other drawings based on the following drawings without creative effort.

[0020] Figure 1 is a flow chart of a method according to an embodiment of the present invention. DETAILED DESCRIPTION

[0021] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only 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 making creative efforts are within the scope of protection of the present invention.

[0022] The present disclosure provides a method for predicting and warning optical fiber faults based on machine learning. Figure 1 As shown, the following steps are included: S1. Acquire multi-dimensional optical characteristic data of the optical fiber; S2. marking the fault area of ​​the multi-dimensional optical feature data to obtain marked feature data; S3, preprocessing the annotated feature data to construct a standard optical fiber fault feature data set; S4, inputting the time-series optical feature data in the multidimensional optical feature data into a neural network model, obtaining the annotated feature data of each time period in the time-series optical feature and the predicted annotated feature data of the future time period, and correcting them with a standard optical fiber fault feature data set; S5. Input the multi-dimensional optical characteristic data sampled in real time from the optical fiber to be tested into the neural network model of S4, and output the optical fiber fault data and prediction data.

[0023] As a preferred embodiment, the multi-dimensional optical feature data includes at least deformation features and transmission features; The deformation characteristics are determined according to the surface curvature of the optical fiber, comprising the following steps: Obtaining the surface curvature of each point on the surface and inside of the optical fiber, and determining whether the surface curvature at each point is within a preset threshold range; If it is not within the preset threshold range, obtaining the surface curvature of the adjacent points of the current point until obtaining a point whose surface curvature is within the preset threshold range; Acquire multiple abnormal deformation parts formed by the point set to obtain the deformation feature; The transmission characteristics are determined according to the transmission parameters of the optical fiber, and include the following steps: Acquire an optical signal transmitted from the deformable feature portion to another deformable feature portion or a non-deformable feature portion, and determine a current optical attenuation coefficient; Obtaining a light attenuation coefficient sequence, wherein each unit of the sequence is set to the light attenuation coefficient and transmission distance from the current deformation feature portion to one of the deformation feature portions or the non-deformation feature portion; Obtaining a light attenuation coefficient set, wherein the light attenuation coefficient set is configured to include a sequence of light attenuation coefficients of each of the deformation feature portions; The transmission characteristics are obtained by comparing each unit of the light attenuation coefficient set with the light attenuation coefficient threshold interval preset in the transmission parameters.

[0024] As a preferred method, marking the fault area in step S2 includes the following steps: When the surface curvature deviation of the abnormally deformed part exceeds the preset threshold and the corresponding light attenuation coefficient exceeds the threshold range but does not fluctuate significantly, it is marked as the first fault type; When the difference in the light attenuation coefficients of adjacent units in the light attenuation coefficient sequence exceeds the preset fluctuation threshold, and the surface curvature deviation of the abnormal deformation part is within the allowable range, it is marked as the second fault type; When the light attenuation coefficients corresponding to the abnormal deformation parts exceeding the preset ratio all exceed the threshold range, and the light attenuation coefficients corresponding to the transmission distance are generally high, it is marked as the third fault type.

[0025] As a preferred embodiment, the pretreatment includes: Eliminate coordinate points in deformation characteristics that exceed the physical size range of the optical fiber, and abnormal optical attenuation coefficients outside the distortion range in transmission characteristics; The surface curvature value in the deformation feature is normalized to the interval [0, 1] based on the curvature threshold of the standard optical fiber, and the optical attenuation coefficient in the transmission feature is normalized based on the theoretical optimal attenuation rate; A mapping relationship between the coordinates of the abnormal deformation part and the corresponding light attenuation coefficient sequence is established to form an associated feature vector.

[0026] As a preferred embodiment, the neural network model in step S4 includes: The time series feature extraction layer uses a long short-term memory network to extract time-related features from the time series optical feature data, including the time series data of the optical attenuation coefficient sequence changes and the position expansion time series data of the deformation abnormal part within a continuous time period; The spatial feature extraction layer uses a convolutional neural network to extract the spatial distribution features of the three-dimensional coordinates of the abnormal deformation part in the deformation feature; The fusion prediction layer fuses the outputs of the temporal feature extraction layer and the spatial feature extraction layer, and outputs the labeled feature data of each time period and the predicted labeled feature data of the future time period.

[0027] As a preferred embodiment, the correction using the standard optical fiber fault feature data set in step S4 includes the following steps: Compare the predicted and marked feature data with the features of the same type of fault in the standard optical fiber fault feature data set, and calculate the feature deviation value; If the deviation value exceeds the preset correction threshold, the weight parameters of the neural network model are adjusted through the back propagation algorithm, and the correction is repeated until the deviation value is within the preset allowable range.

[0028] As an optimal method, when outputting prediction data, the impact of the expansion of the deformation area on the transmission coefficient is simulated, the three-dimensional coordinate expansion range of the abnormal deformation part and the corresponding transmission coefficient change value within the future preset time period are predicted, and the prediction results of the fault development trend are output.

[0029] Specifically, S1. Acquire multi-dimensional optical characteristic data of an optical fiber, where the multi-dimensional optical characteristic data includes deformation characteristics and transmission characteristics.

[0030] To obtain deformation features, relevant detection equipment is used to obtain the surface curvature of each point on the surface and inside the optical fiber, with a preset surface curvature threshold interval of [R1, R2]. For each point, its surface curvature is determined to be within [R1, R2]. If not, the surface curvature of the adjacent points is obtained until a point with surface curvature within [R1, R2] is found. The collection of these points is considered a deformation anomaly, and multiple deformation anomalies together constitute the deformation feature.

[0031] To obtain the transmission signature, the optical signal transmission from each deformed feature to other deformed features or non-deformed features is first acquired. The current optical attenuation coefficient is calculated based on the optical signal transmission. Next, an optical attenuation coefficient sequence is constructed, with each cell in the sequence containing the optical attenuation coefficient and transmission distance from the current deformed feature to another deformed feature or non-deformed feature. The optical attenuation coefficient sequences of all deformed features are combined into an optical attenuation coefficient set. Each cell in the optical attenuation coefficient set is compared with a preset optical attenuation coefficient threshold range [K1, K2] to obtain the transmission signature.

[0032] S2. Mark the fault area of ​​the multi-dimensional optical feature data, preset the surface curvature deviation threshold of the abnormal deformation part to ΔR, the light attenuation coefficient threshold interval to [K1, K2], the light attenuation coefficient difference preset fluctuation threshold to ΔK, and the preset ratio to 50%.

[0033] When the surface curvature deviation of the abnormally deformed part exceeds ΔR, and the corresponding light attenuation coefficient exceeds [K1, K2] but does not fluctuate significantly, it is marked as the first fault type; When the difference in the light attenuation coefficients of adjacent units in the light attenuation coefficient sequence exceeds ΔK, and the surface curvature deviation of the abnormal deformation part is within the allowable range, it is marked as the second fault type; When the optical attenuation coefficients corresponding to more than 50% of the abnormal deformation parts exceed [K1, K2], and the optical attenuation coefficients corresponding to the transmission distance are generally high, it is marked as the third fault type.

[0034] S3. Preprocess the labeled feature data and construct a standard optical fiber fault feature dataset.

[0035] The preprocessing process is as follows: Eliminate coordinate points in the deformation characteristics that exceed the physical size range of the optical fiber. For example, if the actual length of the optical fiber is L, the points whose length direction exceeds L in the coordinate points will be eliminated. Eliminate abnormal optical attenuation coefficients outside the distortion range in the transmission characteristics, such as optical attenuation coefficients that are negative or far greater than the normal range.

[0036] The surface curvature value in the deformation feature is normalized using the curvature threshold R0 of the standard optical fiber as the benchmark. The normalization formula is: Normalized curvature = (actual curvature - R0) / (Rmax - Rmin), where Rmax and Rmin are the preset maximum and minimum curvature values, respectively, ensuring that the normalized curvature is in the range [0, 1]. The optical attenuation coefficient in the transmission feature is normalized using the theoretical optimal attenuation rate K0 as the benchmark. The normalized optical attenuation coefficient = (actual optical attenuation coefficient - K0) / (Kmax - Kmin), where Kmax and Kmin are the preset maximum and minimum optical attenuation coefficient values, respectively.

[0037] A mapping relationship between the three-dimensional coordinates of the abnormal deformation part and the corresponding light attenuation coefficient sequence is established to form an associated feature vector. For example, for the abnormal deformation part A (x1, y1, z1), its corresponding light attenuation coefficient sequence is [K11, d11; K12, d12; ...; K1n, d1n], then the associated feature vector is (x1, y1, z1, K11, d11, K12, d12, ..., K1n, d1n).

[0038] S4. Input the time series optical feature data into the neural network model and perform correction The neural network model includes a temporal feature extraction layer, a spatial feature extraction layer and a fusion prediction layer.

[0039] The time series feature extraction layer uses a long short-term memory network (LSTM) to input the time series data of the sequence changes of the light attenuation coefficient and the position of the abnormal deformation part in a continuous time period. For example, data is collected once an hour, and 100 hours of data are collected as input to extract the time correlation features.

[0040] The spatial feature extraction layer uses a convolutional neural network (CNN) to extract the spatial distribution features of the three-dimensional coordinates of the abnormal deformation part in the deformation feature, and obtains the spatial features through convolution and pooling operations.

[0041] The fusion prediction layer fuses the outputs of the temporal feature extraction layer and the spatial feature extraction layer, for example, by splicing and processing the two features through a fully connected layer, and outputs the labeled feature data of each time period and the predicted labeled feature data of the future time period.

[0042] Correction process: The predicted and annotated feature data is compared with the features of the same type of fault in the standard fiber fault feature dataset, and the feature deviation value is calculated. For example, for the first fault type, the difference between the predicted surface curvature deviation and the surface curvature deviation in the standard data, as well as the difference between the predicted optical attenuation coefficient and the optical attenuation coefficient in the standard data, are calculated to obtain the feature deviation value. The preset correction threshold is 0.1. If the deviation value exceeds 0.1, the weight parameters of the neural network model are adjusted through the backpropagation algorithm. The training and correction process is repeated until the deviation value is within 0.1.

[0043] S5. Fault prediction and early warning for the optical fiber under test The optical fiber to be tested is sampled in real time to obtain its multi-dimensional optical characteristic data, which is pre-processed according to the method of step S3 and then input into the neural network model trained in step S4.

[0044] The model outputs fiber fault data, indicating whether a fault exists in the current fiber and the type of fault. It also outputs prediction data, simulating the impact of deformation zone expansion on the transmission coefficient. For example, assuming the deformation anomaly expands at a certain rate, the model calculates the three-dimensional coordinate expansion range at different time points and the corresponding transmission coefficient change based on the expansion range. The model then predicts the three-dimensional coordinate expansion range and the corresponding transmission coefficient change over the next 24 hours, outputting a prediction of the fault's development trend. For example, "Within the next 12 hours, the deformation anomaly of the first fault type will expand from (x1, y1, z1) to (x2, y2, z2), and the optical attenuation coefficient will increase from K1 to K2."

[0045] The above description and the accompanying drawings sufficiently illustrate the embodiments of the present disclosure to enable those skilled in the art to practice them. Other embodiments may include structural, logical, electrical, process and other changes. The embodiments represent only possible variations. Unless expressly required, individual components and functions are optional, and the order of operations may vary. Portions and features of some embodiments may be included in or replace portions and features of other embodiments. Moreover, the terms used in this application are only used to describe the embodiments and are not used to limit the claims. As used in the description of the embodiments and claims, the singular forms "a", "an" and "the" are intended to also include the plural forms unless the context clearly indicates otherwise. Similarly, the term "and / or" as used in this application means any and all possible combinations of one or more of the associated listings. In addition, when used in this application, the term "comprise" and its variations "comprises" and / or comprising refer to the presence of stated features, wholes, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or groups thereof. In the absence of further restrictions, an element defined by the sentence "comprising a..." does not exclude the presence of other identical elements in the process, method or device that includes the element. In this article, each embodiment may focus on the differences from other embodiments, and the same and similar parts between the various embodiments can be referenced to each other. For the methods, products, etc. disclosed in the embodiments, if they correspond to the method part disclosed in the embodiments, then the relevant parts can be referred to the description of the method part.

[0046] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these effects are performed in hardware or software may depend on the specific application and design constraints of the technical solution. The technicians may use different methods for each specific application to implement the described effects, but such implementations should not be considered to exceed the scope of the embodiments of the present disclosure. The technicians will clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described devices, devices and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0047] The flowcharts and block diagrams in the accompanying drawings show the possible architectures, functions and operations of the devices, methods and computer program products according to the embodiments of the present disclosure. In this regard, each box in the flowchart or block diagram can represent a module, program segment or part of the code, and the module, program segment or part of the code contains one or more executable instructions for implementing the specified logical functions. In some alternative implementations, the functions marked in the boxes can also occur in an order different from that marked in the accompanying drawings. For example, two consecutive boxes can actually be executed in parallel, or they can sometimes be executed in the opposite order, which can depend on the functions involved. In the descriptions corresponding to the flowcharts and block diagrams in the accompanying drawings, the operations or steps corresponding to different boxes can also occur in an order different from that disclosed in the description, and sometimes there is no specific order between different operations or steps. For example, two consecutive operations or steps can actually be executed in parallel, or they can sometimes be executed in the opposite order, which can depend on the functions involved. Each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, may be implemented by dedicated hardware-based devices that perform the specified functions or actions, or may be implemented by a combination of dedicated hardware and computer instructions.

Claims

1. A fiber optic fault prediction and early warning method based on machine learning, characterized in that: The steps include: S1. Acquire multi-dimensional optical characteristic data of the optical fiber; S2. marking the fault area of ​​the multi-dimensional optical feature data to obtain marked feature data; S3, preprocessing the annotated feature data to construct a standard optical fiber fault feature data set; S4, inputting the time-series optical feature data in the multidimensional optical feature data into a neural network model, obtaining the annotated feature data of each time period in the time-series optical feature and the predicted annotated feature data of the future time period, and correcting them with a standard optical fiber fault feature data set; S5. Input the multi-dimensional optical characteristic data sampled in real time from the optical fiber to be tested into the neural network model of S4, and output the optical fiber fault data and prediction data.

2. The optical fiber fault prediction and early warning method based on machine learning according to claim 1, characterized in that: The multi-dimensional optical feature data at least includes deformation features and transmission features; The deformation characteristics are determined according to the surface curvature of the optical fiber, comprising the following steps: Obtaining the surface curvature of each point on the surface and inside of the optical fiber, and determining whether the surface curvature at each point is within a preset threshold range; If it is not within the preset threshold range, obtaining the surface curvature of the adjacent points of the current point until obtaining a point whose surface curvature is within the preset threshold range; Acquire multiple abnormal deformation parts formed by the point set to obtain the deformation feature; The transmission characteristics are determined according to the transmission parameters of the optical fiber, and include the following steps: Acquire an optical signal transmitted from the deformable feature portion to another deformable feature portion or a non-deformable feature portion, and determine a current optical attenuation coefficient; Obtaining a light attenuation coefficient sequence, wherein each unit of the sequence is set to the light attenuation coefficient and transmission distance from the current deformation feature portion to one of the deformation feature portions or the non-deformation feature portion; Obtaining a light attenuation coefficient set, wherein the light attenuation coefficient set is configured to include a sequence of light attenuation coefficients of each of the deformation feature portions; The transmission characteristics are obtained by comparing each unit of the light attenuation coefficient set with the light attenuation coefficient threshold interval preset in the transmission parameters.

3. The optical fiber fault prediction and early warning method based on machine learning according to claim 2, characterized in that: The step S2 of marking the fault area includes the following steps: When the surface curvature deviation of the abnormally deformed part exceeds the preset threshold and the corresponding light attenuation coefficient exceeds the threshold range but does not fluctuate significantly, it is marked as the first fault type; When the difference in the light attenuation coefficients of adjacent units in the light attenuation coefficient sequence exceeds the preset fluctuation threshold, and the surface curvature deviation of the abnormal deformation part is within the allowable range, it is marked as the second fault type; When the light attenuation coefficients corresponding to the abnormal deformation parts exceeding the preset ratio all exceed the threshold range, and the light attenuation coefficients corresponding to the transmission distance are generally high, it is marked as the third fault type.

4. The optical fiber fault prediction and early warning method based on machine learning according to claim 1, characterized in that: The pretreatment includes: Eliminate coordinate points in deformation characteristics that exceed the physical size range of the optical fiber, and abnormal optical attenuation coefficients outside the distortion range in transmission characteristics; The surface curvature value in the deformation feature is normalized to the interval [0, 1] based on the curvature threshold of the standard optical fiber, and the optical attenuation coefficient in the transmission feature is normalized based on the theoretical optimal attenuation rate; A mapping relationship between the coordinates of the abnormal deformation part and the corresponding light attenuation coefficient sequence is established to form an associated feature vector.

5. The optical fiber fault prediction and early warning method based on machine learning according to claim 4, characterized in that: The neural network model in step S4 includes: The time series feature extraction layer uses a long short-term memory network to extract time-related features from the time series optical feature data, including the time series data of the optical attenuation coefficient sequence changes and the position expansion time series data of the deformation abnormal part within a continuous time period; The spatial feature extraction layer uses a convolutional neural network to extract the spatial distribution features of the three-dimensional coordinates of the abnormal deformation part in the deformation feature; The fusion prediction layer fuses the outputs of the temporal feature extraction layer and the spatial feature extraction layer, and outputs the labeled feature data of each time period and the predicted labeled feature data of the future time period.

6. The optical fiber fault prediction and early warning method based on machine learning according to claim 5, characterized in that: The correction in step S4 is performed using the standard optical fiber fault feature data set, including the following steps: Compare the predicted and marked feature data with the features of the same type of fault in the standard optical fiber fault feature data set, and calculate the feature deviation value; If the deviation value exceeds the preset correction threshold, the weight parameters of the neural network model are adjusted through the back propagation algorithm, and the correction is repeated until the deviation value is within the preset allowable range.

7. The optical fiber fault prediction and early warning method based on machine learning according to claim 6, characterized in that: When outputting prediction data, the impact of deformation area expansion on the transmission coefficient is simulated, the three-dimensional coordinate expansion range of the abnormal deformation part and the corresponding transmission coefficient change value within the future preset time period are predicted, and the prediction results of the fault development trend are output.

Images