A data processing-based ADSS optical cable inspection method and system

Abstract

This invention provides an ADSS optical cable inspection method and system based on data processing. The invention relates to the field of optical cable inspection technology. The method includes acquiring images along the ADSS optical cable line; using a risk distribution model based on the ADSS optical cable line images to determine multiple first-risk optical cable segments and a first fault risk distribution map for each first-risk optical cable segment; constructing a first risk map; processing the first risk map using a graph convolutional network to determine multiple supplementary detection points for the multiple first-risk optical cable segments; and determining cable fault points based on distributed fiber strain sensing data from multiple preliminary detection points of second-risk optical cable segments and distributed fiber strain sensing data from multiple supplementary detection points of second-risk optical cable segments. This method can efficiently identify high-risk points in ADSS optical cables and achieve accurate positioning.

Description

Technical Field

[0001] This invention relates to the field of optical cable testing technology, and specifically to an ADSS optical cable testing method and system based on data processing. Background Technology

[0002] With the rapid development of power communication networks, ADSS optical cables, as a crucial carrier of power system communication, directly impact the safety and stability of power communication due to their operational reliability. However, because ADSS optical cables are exposed to complex and variable environments for extended periods, they are susceptible to potential faults due to factors such as mechanical stress, temperature variations, and wind vibration. Traditional inspection methods primarily rely on manual inspections and periodic testing. This is time-consuming and labor-intensive, not only inefficient but also unable to comprehensively and accurately identify potential risk points in the optical cable. Especially in long-distance, large-scale line inspections, existing technologies often cannot accurately locate high-risk sections, leading to significant delays in fault prevention and maintenance. Furthermore, conventional inspection methods have limited ability to monitor the strain distribution and micro-deformation of optical cables, making it difficult to detect early-stage problems in a timely manner, which can easily cause sudden failures and seriously affect the reliability of power communication systems.

[0003] Therefore, how to efficiently identify high-risk points of ADSS optical cables and achieve accurate positioning is an urgent problem to be solved. Summary of the Invention

[0004] The main technical problem this invention addresses is how to efficiently identify high-risk points in ADSS optical cables and achieve accurate positioning.

[0005] According to a first aspect, the present invention provides an ADSS optical cable inspection method based on data processing, comprising: acquiring images along the ADSS optical cable line; determining multiple first-risk optical cable segments and a first fault risk distribution map for each first-risk optical cable segment using a risk distribution model based on the images along the ADSS optical cable line; determining multiple preliminary detection points for each first-risk optical cable segment based on the images of the multiple first-risk optical cable segments; acquiring distributed fiber strain sensing data of the multiple preliminary detection points for each first-risk optical cable segment; constructing a first risk map, the first risk map including multiple preliminary detection nodes and edges between the multiple nodes, wherein the node features of each preliminary detection node are multiple preliminary detection points for each first-risk optical cable segment. The data includes distributed fiber optic strain sensing data at the detection points, ADSS images along the optical cable line, and the edges between preliminary detection nodes representing their relative positions and distances. A graph convolutional network is used to process the first risk map to determine multiple supplementary detection points for multiple first-risk optical cable segments, and distributed fiber optic strain sensing data for each of these supplementary detection points is acquired. Based on this data, multiple second-risk optical cable segments are determined. Finally, cable fault points are determined based on the distributed fiber optic strain sensing data from multiple preliminary detection points and the distributed fiber optic strain sensing data from multiple supplementary detection points in the second-risk optical cable segments.

[0006] In one possible implementation, determining multiple second-risk optical cable segments based on distributed optical fiber strain sensing data from multiple supplementary detection points of each first-risk optical cable segment includes: generating a second fault risk distribution map for each first-risk optical cable segment based on distributed optical fiber strain sensing data from multiple preliminary detection points of each first-risk optical cable segment and distributed optical fiber strain sensing data from multiple supplementary detection points of each first-risk optical cable segment; and determining multiple second-risk optical cable segments based on the first fault risk distribution map and the second fault risk distribution map of each first-risk optical cable segment.

[0007] In one possible implementation, determining the cable fault point based on distributed fiber optic strain sensing data from multiple preliminary detection points of the second-risk optical cable segment and distributed fiber optic strain sensing data from multiple supplementary detection points of the second-risk optical cable segment includes: acquiring environmental videos of multiple second-risk optical cable segments; determining multiple key inspection points for each second-risk optical cable segment based on the environmental videos of the multiple second-risk optical cable segments, distributed fiber optic strain sensing data from multiple preliminary detection points of the second-risk optical cable segment, and distributed fiber optic strain sensing data from multiple supplementary detection points of the second-risk optical cable segment; and determining the cable fault point based on OTDR test data from the multiple key inspection points of each second-risk optical cable segment.

[0008] In one possible implementation, the risk distribution model is a convolutional neural network model.

[0009] According to a second aspect, the present invention provides an ADSS optical cable inspection system based on data processing, comprising: an acquisition module for acquiring images along the ADSS optical cable line; a risk calculation module for determining multiple first-risk optical cable segments and a first fault risk distribution map for each first-risk optical cable segment based on the images along the ADSS optical cable line using a risk distribution model; a preliminary detection module for determining multiple preliminary detection points for each first-risk optical cable segment based on the images of the multiple first-risk optical cable segments; a first sensing data module for acquiring distributed fiber strain sensing data of the multiple preliminary detection points for each first-risk optical cable segment; and a construction module for constructing a first risk map, the first risk map including multiple preliminary detection nodes and edges between the multiple nodes, wherein the node features of each preliminary detection node are each first-risk optical cable segment. The system comprises: distributed fiber optic strain sensing data from multiple preliminary detection points of the optical cable segment; ADSS images along the optical cable line; and edges between preliminary detection nodes representing their relative positions and distances. A supplementary detection module processes the first risk map using a graph convolutional network to determine multiple supplementary detection points for multiple first-risk optical cable segments and acquires distributed fiber optic strain sensing data from these supplementary detection points for each first-risk optical cable segment. A secondary risk module determines multiple second-risk optical cable segments based on the distributed fiber optic strain sensing data from the supplementary detection points for each first-risk optical cable segment. A repair and location module determines cable fault points based on the distributed fiber optic strain sensing data from multiple preliminary detection points and the distributed fiber optic strain sensing data from the supplementary detection points for each second-risk optical cable segment.

[0010] In one possible implementation, the secondary risk module is further configured to: generate a second fault risk distribution map for each first-risk optical cable segment based on distributed optical fiber strain sensing data from multiple preliminary detection points of each first-risk optical cable segment and distributed optical fiber strain sensing data from multiple supplementary detection points of each first-risk optical cable segment; and determine multiple second-risk optical cable segments based on the first fault risk distribution map and the second fault risk distribution map of each first-risk optical cable segment.

[0011] In one possible implementation, the repair and location module is further configured to: acquire environmental videos of multiple second-risk optical cable segments; determine multiple key inspection points for each second-risk optical cable segment based on the environmental videos of the multiple second-risk optical cable segments, distributed fiber strain sensing data of multiple preliminary detection points of the second-risk optical cable segments, and distributed fiber strain sensing data of multiple supplementary detection points of the second-risk optical cable segments; and determine the cable fault point based on the OTDR test data of the multiple key inspection points of each second-risk optical cable segment.

[0012] In one possible implementation, the risk distribution model is a convolutional neural network model.

[0013] According to a third aspect, embodiments of the present invention provide an electronic device, including: a processor; a memory; and a computer program; wherein the computer program is stored in the memory and configured to be executed by the processor to implement the method as described above, the method comprising: acquiring an image along an ADSS optical cable line; determining, based on the image along the ADSS optical cable line, a plurality of first-risk optical cable segments and a first fault risk distribution map for each first-risk optical cable segment using a risk distribution model; determining, based on the images of the plurality of first-risk optical cable segments, a plurality of preliminary detection points for each first-risk optical cable segment; acquiring distributed fiber strain sensing data of the plurality of preliminary detection points for each first-risk optical cable segment; and constructing a first risk map, the first risk map including a plurality of preliminary detection nodes and edges between the plurality of nodes. Each preliminary detection node is characterized by distributed fiber strain sensing data of multiple preliminary detection points for each first-risk optical cable segment, and an ADSS image along the optical cable line. The edges between preliminary detection nodes represent their relative positional relationships and distances. The first risk map is processed using a graph convolutional network to determine multiple supplementary detection points for multiple first-risk optical cable segments, and distributed fiber strain sensing data of these supplementary detection points for each first-risk optical cable segment is acquired. Multiple second-risk optical cable segments are determined based on the distributed fiber strain sensing data of the multiple supplementary detection points for each first-risk optical cable segment. Cable fault points are determined based on the distributed fiber strain sensing data of the multiple preliminary detection points and the distributed fiber strain sensing data of the multiple supplementary detection points for the second-risk optical cable segments.

[0014] According to the fourth aspect, this embodiment provides a computer-readable storage medium storing a computer program thereon. When executed by a processor, the program implements the aforementioned ADSS optical cable detection method based on data processing. The method includes: acquiring an image along the ADSS optical cable line; determining multiple first-risk optical cable segments and a first fault risk distribution map for each first-risk optical cable segment using a risk distribution model based on the image along the ADSS optical cable line; determining multiple preliminary detection points for each first-risk optical cable segment based on the images of the multiple first-risk optical cable segments; acquiring distributed fiber strain sensing data of the multiple preliminary detection points for each first-risk optical cable segment; and constructing a first risk map, the first risk map including multiple preliminary detection nodes and edges between the multiple nodes, each preliminary detection... The node characteristics are distributed fiber strain sensing data of multiple preliminary detection points for each first-risk optical cable segment, and ADSS optical cable line images. The edges between preliminary detection nodes represent their relative positional relationships and distances. The first risk map is processed using a graph convolutional network to determine multiple supplementary detection points for multiple first-risk optical cable segments, and distributed fiber strain sensing data of these supplementary detection points for each first-risk optical cable segment is acquired. Multiple second-risk optical cable segments are determined based on the distributed fiber strain sensing data of these supplementary detection points for each first-risk optical cable segment. Cable fault points are determined based on the distributed fiber strain sensing data of the multiple preliminary detection points and the distributed fiber strain sensing data of the multiple supplementary detection points for the second-risk optical cable segments.

[0015] This invention provides an ADSS optical cable inspection method and system based on data processing. The method includes acquiring images along the ADSS optical cable line; determining multiple first-risk optical cable segments and a first fault risk distribution map for each first-risk optical cable segment using a risk distribution model based on the images along the ADSS optical cable line; determining multiple preliminary detection points for each first-risk optical cable segment based on the images of the multiple first-risk optical cable segments; acquiring distributed fiber strain sensing data of the multiple preliminary detection points for each first-risk optical cable segment; and constructing a first risk map, the first risk map including multiple preliminary detection nodes and edges between the nodes, wherein the node feature of each preliminary detection node is the distributed fiber strain sensing data of the multiple preliminary detection points for each first-risk optical cable segment. Based on the images along the ADSS optical cable line, the edges between the preliminary detection nodes represent the relative positional relationship and distance between them. A graph convolutional network is used to process the first risk map to determine multiple supplementary detection points for multiple first-risk optical cable segments, and distributed fiber strain sensing data for each first-risk optical cable segment's multiple supplementary detection points are acquired. Based on the distributed fiber strain sensing data for each first-risk optical cable segment's multiple supplementary detection points, multiple second-risk optical cable segments are determined. Based on the distributed fiber strain sensing data for the multiple preliminary detection points of the second-risk optical cable segments and the distributed fiber strain sensing data for the multiple supplementary detection points of the second-risk optical cable segments, cable fault points are determined. This method can efficiently identify high-risk points in ADSS optical cables and achieve accurate positioning. Attached Figure Description

[0016] Figure 1 A flowchart illustrating an ADSS optical cable detection method based on data processing, provided in an embodiment of the present invention;

[0017] Figure 2 A schematic diagram of an ADSS optical cable provided in an embodiment of the present invention;

[0018] Figure 3 A schematic diagram of a process for determining multiple second risk optical cable segments provided in an embodiment of the present invention;

[0019] Figure 4 This is a schematic diagram of a process for determining a cable fault point according to an embodiment of the present invention;

[0020] Figure 5 This is a schematic diagram of an ADSS optical cable detection system based on data processing, provided as an embodiment of the present invention. Detailed Implementation

[0021] The present invention will now be described in further detail with reference to specific embodiments and accompanying drawings. Similar elements in different embodiments are referred to by associated similar element reference numerals. In the following embodiments, many details are described to facilitate a better understanding of the invention. However, those skilled in the art will readily recognize that some features may be omitted in different situations, or may be replaced by other elements, materials, or methods. In some cases, certain operations related to the present invention are not shown or described in the specification. This is to avoid obscuring the core parts of the invention with excessive description. For those skilled in the art, detailed description of these related operations is not necessary; they can fully understand the related operations based on the description in the specification and general technical knowledge in the art.

[0022] In this embodiment of the invention, the following are provided: Figure 1 The method for detecting ADSS optical cables based on data processing is shown, and the method includes steps S1 to S8:

[0023] Step S1: Obtain images along the ADSS optical cable line.

[0024] ADSS optical cable, or all-dielectric self-supporting optical cable, is a communication optical cable that contains no metal components and relies on its own structural strength for self-support and suspension. ADSS optical cable can be used in complex environments such as high-voltage transmission lines to achieve data transmission. Figure 2 This is a schematic diagram of an ADSS optical cable provided in an embodiment of the present invention.

[0025] The images along the ADSS optical cable line are images of the ADSS optical cable itself captured by imaging equipment installed along the ADSS optical cable line.

[0026] The images along the ADSS optical cable line record the appearance of the ADSS optical cable line, such as the cable's direction, surface condition, and joint location.

[0027] Step S2: Based on the images along the ADSS optical cable line, a risk distribution model is used to determine multiple first-risk optical cable segments and a first fault risk distribution map for each first-risk optical cable segment.

[0028] The risk distribution model is a convolutional neural network model. The input to the risk distribution model is the image along the ADSS optical cable line, and the output of the risk distribution model is multiple first-risk optical cable segments and a first fault risk distribution map for each first-risk optical cable segment.

[0029] Convolutional Neural Network (CNN) models are a type of deep learning model used for image processing. A CNN consists of convolutional layers, pooling layers, and fully connected layers. Convolutional layers extract local features from an image using convolutional kernels, pooling layers reduce feature dimensionality while preserving key information, and fully connected layers synthesize the extracted features to output classification or prediction results.

[0030] The multiple first-risk optical cable segments were selected by analyzing the images along the ADSS optical cable line using a risk distribution model, identifying ADSS optical cable line segments with a certain probability of failure.

[0031] The first risky optical cable segment exhibits characteristics that could lead to failure, such as surface wear and abnormal suspension.

[0032] The first fault risk distribution map for each first-risk optical cable segment is a distribution map generated by a risk distribution model to represent the probability of faults occurring at different locations on each first-risk optical cable segment.

[0033] The first fault risk distribution map uses different values ​​and colors to visually represent the fault risk level at each location on the ADSS optical cable.

[0034] Images along the ADSS optical cable line contain the cable's own appearance features, which implicitly contain visual cues that may lead to faults, such as color changes due to cable aging and minor surface damage. These features provide the basic data needed for risk distribution models to identify risky cable segments and generate risk distribution maps. Convolutional neural networks can automatically learn fault-related visual features from ADSS optical cable line images, enabling preliminary identification of optical cable risks.

[0035] Convolutional neural networks (CNNs) can perform preprocessing operations such as cropping and normalization on images along the ADSS optical cable line. Then, convolutional layers extract local features of the optical cable, such as surface defects and morphological anomalies. Pooling layers compress these extracted features, reducing computation while preserving key features. Subsequently, fully connected layers compare the extracted features with known optical cable fault sample features to calculate the fault probability of each cable segment. The final model outputs multiple first-risk optical cable segments and generates a first fault risk distribution map for each segment based on the fault probability at each location.

[0036] Step S3: Determine multiple preliminary detection points for each first-risk optical cable segment based on images of multiple first-risk optical cable segments.

[0037] The images of multiple first-risk optical cable segments are the images corresponding to multiple first-risk optical cable segment portions in the images along the ADSS optical cable line.

[0038] In some embodiments, a detection and localization model can be used to determine multiple preliminary detection points for each first-risk optical cable segment. The detection and localization model is a convolutional neural network. The input to the detection and localization model is an image of multiple first-risk optical cable segments, and the output of the detection and localization model is multiple preliminary detection points for each first-risk optical cable segment.

[0039] The multiple preliminary detection points for each first-risk optical cable segment are the specific locations determined by the detection and positioning model that require further sensor data collection.

[0040] The images of the first-risk fiber optic cable segment contain detailed visual information. Anomalies within these segments directly indicate potential fault locations, providing analyzable visual evidence for the model to determine initial detection points and enabling precise location of areas requiring focused monitoring. The convolutional neural network focuses on detailed features within multiple images of the first-risk fiber optic cable segments. Convolutional layers further extract local anomalies within each segment, such as micro-cracks on the cable surface and locations of contact with other objects. Pooling layers enhance the saliency of these anomalies. The model can then mark locations with high feature matching as initial detection points. Ultimately, the model outputs multiple initial detection points for each first-risk fiber optic cable segment, covering the most likely critical locations within that segment where potential faults may exist.

[0041] In some embodiments, determining multiple preliminary detection points for each first-risk optical cable segment based on images of multiple first-risk optical cable segments includes steps S31 to S33:

[0042] Step S31: Based on images of multiple first-risk optical cable segments, determine the locations of surface wear points, abnormal suspension distribution trajectories, and the range of surface crack areas on the optical cable surface.

[0043] In some embodiments, a convolutional neural network can be used to determine the location of wear points on the surface of the optical cable, the trajectory of abnormal suspension distribution, and the range of surface crack areas.

[0044] The wear points on the optical cable surface were identified from the image of the first-risk optical cable segment using a convolutional neural network, pinpointing the specific locations of wear marks on the cable surface. These wear points were marked with image coordinates.

[0045] The abnormal suspension distribution trajectory is a continuous positional change path that reflects the abnormal suspension status (including sag and offset) of the optical cable, output by a convolutional neural network.

[0046] The surface crack region is defined by the boundary of the cracked area on the optical cable surface, as output by the convolutional neural network. The surface crack region is marked with an outline to indicate the size and distribution of the crack.

[0047] Convolutional neural networks can preprocess images of multiple first-risk optical cable segments, then extract local features related to wear, suspension anomalies, and cracks, such as grayscale changes and edge contours, through convolutional layers. Pooling layers then compress the features while retaining key information. Finally, fully connected layers classify the features and output the locations of wear points, suspension anomaly distribution trajectories, and surface crack areas on the optical cable surface.

[0048] Step S32: Based on the wear points on the optical cable surface, the abnormal distribution trajectory of the suspension, and the range of the surface crack area, determine multiple risk feature areas, the failure probability score of each risk feature area, and the influence correlation value between risk areas.

[0049] In some embodiments, a deep neural network can be used to determine multiple risk feature regions, the failure probability score of each risk feature region, and the influence correlation value between risk regions.

[0050] Deep Neural Networks (DNNs) are artificial neural network models that contain multiple hidden layers. By mimicking the connection patterns of neurons in the human brain, DNNs can learn features and recognize patterns from complex data. Through multi-layer nonlinear transformations, DNNs can automatically extract low-order to high-order abstract features from input data.

[0051] Multiple risk feature zones are formed by dividing areas with similar characteristics such as surface wear, abnormal suspension, and cracks in optical cables into several continuous zones using deep neural networks. Each zone represents one or more types of risk features.

[0052] The failure probability score for each risk feature area is a quantitative assessment of the probability of failure occurring in each risk feature area, output by a deep neural network. The higher the score, the higher the failure risk.

[0053] The impact correlation value between risk zones is reflected by the output of a deep neural network, indicating the degree of mutual influence between zones with different risk characteristics. The higher the value, the stronger the correlation between the failure risks between zones.

[0054] After receiving data on the wear points, abnormal suspension distribution trajectories, and surface crack areas of the optical cable, the deep neural network first maps these features into structured data. Then, it learns the nonlinear relationships between features through multiple hidden layers. The first layer integrates similar features to form preliminary regions, while the intermediate layers calculate the failure probability of each region as a failure likelihood score. The deep network of the model can analyze the spatial correlation and feature dependencies between regions, ultimately outputting multiple risk feature regions, corresponding failure likelihood scores, and influence correlation values ​​between regions.

[0055] Step S33: Based on the multiple risk feature areas, the fault probability score of each risk feature area, and the influence correlation value between the risk areas, determine multiple preliminary detection points for each first risk optical cable segment.

[0056] In some embodiments, a deep neural network may be used to determine multiple preliminary detection points for each first risky fiber optic cable segment.

[0057] After receiving multiple risk feature regions, the failure probability score of each region, and the influence correlation values ​​between regions, the deep neural network first transforms this data into a structured form that the model can process through the input layer. The first hidden layer standardizes the failure probability scores of each region, such as mapping them to a score range of 0-100, and constructs a region correlation matrix based on the influence correlation values, clarifying the cluster relationship of highly correlated regions. For example, if regions A and B have high correlation values, they are considered as a risk-linked cluster. The intermediate layer performs a secondary calculation on the region scores using a weighted algorithm. For a single high-risk region without strong correlation, its own score can be used as the core weight. For regions within a correlation cluster, the product of the highest score and the average correlation value within the cluster is used as the comprehensive weight to strengthen the impact of the risk linkage effect. The deep network can combine weight ranking to further refine the selection within each high-weight region. The deep network can prioritize the locations with the most significant features within a region, such as dense areas of wear points, crack initiation points, and inflection points of abnormal suspension trajectories, ensuring that the selected locations are spatially evenly distributed to cover the core risk range of the region. Ultimately, the model can output multiple preliminary detection points that have undergone multi-layer screening and spatial optimization to ensure key coverage of high-risk areas, and avoid missing potential linkage risk points through correlation analysis.

[0058] Step S4: Obtain distributed fiber strain sensing data from multiple preliminary detection points in each first-risk fiber optic cable segment.

[0059] Distributed fiber optic strain sensing data from multiple preliminary detection points in each first-risk fiber optic cable segment were collected by distributed fiber optic sensing devices at these points. This data reflects the quantitative information about strain, such as tension and bending deformation, experienced by the cable at these locations. The distributed fiber optic strain sensing data includes information such as strain values ​​and frequency of change.

[0060] Distributed fiber optic strain sensing data can intuitively reflect the stress state of optical cables.

[0061] Step S5: Construct a first risk map. The first risk map includes multiple preliminary detection nodes and edges between the nodes. The node features of each preliminary detection node are distributed fiber strain sensing data and ADSS optical cable line images of multiple preliminary detection points of each first risk optical cable segment. The edges between the preliminary detection nodes are the relative positional relationships and distances between the preliminary detection nodes.

[0062] The first risk map is a map structure used to represent the correlation between multiple preliminary detection points.

[0063] Multiple preliminary detection nodes correspond to multiple preliminary detection points for each first-risk optical cable segment.

[0064] The node features of each preliminary detection node integrate the distributed fiber strain sensing data of that detection point with the ADSS optical cable line image.

[0065] The edges between the initial detection nodes quantify the spatial connections between different detection points through relative positional relationships and distances. The edges of the first risk map can reflect the characteristics and spatial relationships of each detection point.

[0066] Step S6: Process the first risk map based on graph convolutional network to determine multiple supplementary detection points for multiple first risk optical cable segments, and obtain distributed optical fiber strain sensing data of multiple supplementary detection points for each first risk optical cable segment.

[0067] The input to the graph convolutional network is the first risk map, and the output of the graph convolutional network is multiple supplementary detection points of multiple first risk optical cable segments.

[0068] Graph Convolutional Networks (GCNs) are deep learning models suitable for graphs. GCNs update node representations by aggregating the features of a node itself and those of its neighboring nodes. GCNs effectively uncover potential relationships between nodes in a graph and can capture global structural information using these connections. GCNs exhibit strong performance when processing data with complex relationships.

[0069] Multiple supplementary detection points for several first-risk optical cable segments were identified as locations requiring additional sensor data collection after processing the first-risk map using a graph convolutional network. These locations are areas not initially covered by the detection points but spatially related to high-risk nodes, as well as potential risk locations inferred through node feature correlation.

[0070] By constructing a first risk map, the spatial relationships between multiple preliminary detection nodes can be clearly reflected. The fault risks of optical cables often exhibit spatial correlation; an abnormal state at one detection point may influence surrounding detection points. Therefore, this correlation information is crucial for determining supplementary detection points. Using the distributed fiber strain sensing data of each preliminary detection node and the ADSS optical cable line image as node features, while using the relative positional relationships and distances between preliminary detection nodes as edge features, allows for more comprehensive utilization of data information. This helps the graph convolutional network better understand the state correlations and spatial influences of each preliminary detection point, thereby improving the comprehensiveness of supplementary detection point location. Processing the first risk map using a graph convolutional network can effectively learn the complex relationships and information transmission between nodes, thus more accurately uncovering potential risk areas not covered by the preliminary detection points. Compared to traditional methods of isolated detection point analysis, graph convolutional networks have stronger global correlation capture and risk prediction capabilities when processing this type of graph-structured data.

[0071] When processing the first risk map, the graph convolutional network first initializes the node features in the first risk map. The graph convolutional network can convert the distributed fiber strain sensing data of each preliminary detection node and the image features along the ADSS optical cable line into vector form. Then, the graph convolutional network can aggregate the features of each node and its neighboring nodes through graph convolutional layers, and combine the relative position and distance information of the edges to calculate the global feature representation of the nodes. Through this process, it can capture the spatial correlation and feature dependence between nodes, such as the possibility that high strain at a certain detection point may affect the surrounding area. Through multi-layer graph convolutional operations, the model can gradually strengthen its ability to identify potential risk areas, and finally output the locations with high risk that are not covered by the preliminary detection points, i.e., multiple supplementary detection points for multiple first-risk optical cable segments.

[0072] The distributed fiber optic strain sensing data for each first-risk fiber optic cable segment is the strain data collected by distributed fiber optic sensing devices at multiple supplementary detection points in multiple first-risk fiber optic cable segments.

[0073] Step S7: Determine multiple second-risk optical cable segments based on distributed optical fiber strain sensing data from multiple supplementary detection points of each first-risk optical cable segment.

[0074] In some embodiments, Figure 3 This is a schematic flowchart illustrating the process of determining multiple second-risk optical cable segments according to an embodiment of the present invention. The determination of multiple second-risk optical cable segments includes steps S71-S72:

[0075] Step S71: Based on the distributed fiber strain sensing data of multiple preliminary detection points of each first-risk optical cable segment and the distributed fiber strain sensing data of multiple supplementary detection points of each first-risk optical cable segment, generate a second fault risk distribution map for each first-risk optical cable segment.

[0076] In some embodiments, a second risk distribution model can be used to generate a second fault risk distribution map for each first-risk optical cable segment. The second risk distribution model is a generative adversarial network (GAN). The inputs to the second risk distribution model are distributed fiber strain sensing data from multiple preliminary detection points and multiple supplementary detection points of each first-risk optical cable segment. The output of the second risk distribution model is a second fault risk distribution map for each first-risk optical cable segment.

[0077] Generative Adversarial Networks (GANs) are deep learning models consisting of a generator and a discriminator. The generator learns the distribution of the input data to generate new data samples, while the discriminator is responsible for distinguishing the generated data from the real data. Through adversarial training, the generator and discriminator are continuously optimized, enabling the generator to generate data that closely approximates the real distribution.

[0078] The second fault risk distribution map for each first-risk optical cable segment is generated by the second risk distribution model based on the distributed optical fiber strain sensing data of multiple preliminary detection points and multiple supplementary detection points of each first-risk optical cable segment. It is used to represent the distribution map of fault risk at different locations on the first-risk optical cable segment.

[0079] Compared with the first fault risk distribution map, the second fault risk distribution map is based on more comprehensive sensor data. Therefore, the risk range in the second fault risk distribution map is more accurate, and the risk measurement is more precise.

[0080] The distributed fiber strain sensing data from multiple preliminary detection points and multiple supplementary detection points of each first-risk fiber optic cable segment cover the stress state information at key locations on that segment. This data can reflect the structural health of the fiber optic cable; for example, high strain may indicate a risk of breakage. This provides direct physical quantitative basis for generating adversarial networks to learn the risk distribution pattern and generate a more accurate second fault risk distribution map.

[0081] The generator in a generative adversarial network (GAN) receives distributed fiber strain sensing data from preliminary and supplementary detection points on each first-risk fiber optic cable segment and learns the spatial distribution patterns of this data along the cable segment, such as the variation trend of strain values ​​with location. Based on the learned patterns, the generator can generate a fault risk distribution prediction map covering the entire first-risk fiber optic cable segment. The discriminator then compares the generated prediction map with a real risk distribution constructed from high-risk samples based on the sensor data distribution of known historical fault points, thereby determining the authenticity of the prediction map. The generator adjusts its parameters based on feedback from the discriminator and continuously optimizes the accuracy of the prediction map, enabling the generated second fault risk distribution map to both match existing sensor data and reasonably infer the risk level of undetected locations.

[0082] Step S72: Based on the first fault risk distribution map and the second fault risk distribution map of each first risk optical cable segment, determine a plurality of second risk optical cable segments.

[0083] In some embodiments, a segment analysis model can be used to determine multiple second-risk optical cable segments. The segment analysis model is a Transformer model. The inputs to the segment analysis model are a first fault risk distribution map and a second fault risk distribution map for each first-risk optical cable segment, and the output of the segment analysis model is multiple second-risk optical cable segments.

[0084] The Transformer model consists of an encoder and a decoder. The encoder captures long-range dependencies in the input sequence through a self-attention mechanism and transforms the input data into a feature representation containing global information. The decoder then generates the target sequence based on the encoder's output. The Transformer model can effectively handle data with temporal or spatial relationships.

[0085] Multiple second-risk optical cable segments were identified by analyzing the first and second fault risk distribution maps of each first-risk optical cable segment using a segment analysis model. These segments had higher fault risk and a more precise range.

[0086] The second-risk optical cable segment shows a high degree of consistency in risk in both the first and second fault risk distribution maps, making it a key area for investigation.

[0087] The first and second fault risk distribution maps of each first-risk optical cable segment reflect the risk characteristics of the segment from different dimensions. The differences and consistency between the two maps can reveal the true high-risk areas. For example, if a certain area is high-risk in both maps, then the probability of failure in that area is higher. This provides complementary risk assessment basis for the Transformer model to determine more accurate second-risk optical cable segments.

[0088] The Transformer model can convert the first and second fault risk distribution maps of each first-risk optical cable segment into sequence data, where each element represents the risk value at a certain location on the segment. The encoder can calculate the association weights of each location in the sequence using a self-attention mechanism, such as whether a high risk at a certain location is related to adjacent locations. Then, it integrates the features of the first and second fault risk distribution maps to generate a feature sequence containing global risk associations. Based on this feature sequence and combined with a preset risk threshold, the decoder can identify regions where the risk level in both distribution maps exceeds the threshold and is continuously distributed. Ultimately, these regions can be identified by the model as multiple second-risk optical cable segments.

[0089] Step S8: Determine the cable fault point based on the distributed fiber optic strain sensing data of multiple preliminary detection points of the second risk optical cable segment and the distributed fiber optic strain sensing data of multiple supplementary detection points of the second risk optical cable segment.

[0090] In some embodiments, Figure 4 This is a flowchart illustrating a method for determining a cable fault point according to an embodiment of the present invention. The method for determining the cable fault point includes steps S81-S83:

[0091] Step S81: Obtain environmental videos of multiple second-risk optical cable segments.

[0092] The environmental videos of multiple secondary risk fiber optic cable segments are dynamic videos recorded by high-definition cameras around these segments, documenting the real-time environmental conditions. The videos include dynamic information such as weather, the influence of surrounding objects, and the interaction between the fiber optic cables.

[0093] Step S82: Based on the environmental video of the multiple second-risk optical cable segments, the distributed optical fiber strain sensing data of multiple preliminary detection points of the second-risk optical cable segments, and the distributed optical fiber strain sensing data of multiple supplementary detection points of the second-risk optical cable segments, determine multiple key investigation points for each second-risk optical cable segment.

[0094] In some embodiments, a key location model can be used to determine multiple key inspection points for each second-risk optical cable segment. The key location model is a Transformer model. The inputs to the key location model are environmental video of the multiple second-risk optical cable segments, distributed fiber strain sensing data from multiple preliminary detection points of the second-risk optical cable segments, and distributed fiber strain sensing data from multiple supplementary detection points of the second-risk optical cable segments. The output of the key location model is multiple key inspection points for each second-risk optical cable segment.

[0095] Multiple key inspection points for each second-risk optical cable segment are specific locations identified through a key location model that require final fault confirmation. These locations are overlapping points with significant environmental impact and abnormal sensor data, serving as candidate locations for cable faults.

[0096] Environmental videos from multiple second-risk fiber optic cable segments reflect the dynamic impact of environmental factors on the cable, such as cable swaying caused by wind and friction from tree branches. Distributed fiber optic strain sensing data from multiple preliminary and supplementary detection points in the second-risk fiber optic cable segments quantify the stress state of the cable, enabling the location of high-risk locations under the combined effects of environmental factors and stress.

[0097] The Transformer can convert environmental video of the second-risk optical cable segment into frame sequence features, and convert distributed optical fiber strain sensing data from multiple initial detection points and supplementary detection points of the second-risk optical cable segment into temporal features reflecting strain change trends. The encoder can associate the environmental frame sequence with the sensing temporal features through a self-attention mechanism to capture the causal relationship between environmental factors and optical cable stress. Based on these associated features, the decoder can filter out locations with significant environmental impact and persistently abnormal sensing data. These locations are identified as multiple key investigation points for each second-risk optical cable segment.

[0098] Step S83: Based on the OTDR test data of multiple key inspection points in each second risk optical cable segment, determine the cable fault point.

[0099] The OTDR test data for multiple key inspection points in each second-risk optical cable segment are quantitative data obtained by testing multiple key inspection points in each second-risk optical cable segment using an OTDR (Optical Time Domain Reflectometer). This data records the reflection, scattering, and attenuation characteristics of the optical signal during transmission within the optical cable. The OTDR test data for multiple key inspection points in the second-risk optical cable segment includes information such as reflection peak intensity, attenuation coefficient, and event point distance, such as the distance between the fault point and the test end.

[0100] In some embodiments, a high-risk location model can be used to determine cable fault points based on OTDR test data from multiple key inspection points in each second-risk optical cable segment. The high-risk location model is a deep neural network model. The input to the high-risk location model is OTDR test data from multiple key inspection points in each second-risk optical cable segment, and the output of the high-risk location model is the cable fault point.

[0101] Cable fault points refer to specific locations in each second-risk optical cable segment that, as identified by the high-risk location model, have a risk level exceeding a preset threshold, exhibit significant fault characteristics, and require priority handling.

[0102] Cable fault points correspond to substantial damage to the optical cable and potential for deterioration in the short term. They are also key points that require priority repair work during optical cable maintenance.

[0103] OTDR test data from multiple key inspection points in the second-risk optical cable segment directly reflects physical damage such as cable breakage and excessive bending, as well as performance degradation due to excessive signal attenuation at these key inspection points and surrounding areas. The data contains features such as reflection peaks and attenuation values, which directly correspond to the types of optical cable faults, such as breakage, loose joints, and attenuation caused by wear. It also includes the severity of each fault type, thus providing the model with physical signals to determine the risk level. Through this data, the model can quantitatively analyze the fault status of each key inspection point. Deep neural networks can handle the nonlinear relationships and complex fault characteristics in OTDR test data. Through the collaborative computation of multiple layers of neurons in a deep neural network, precise location of optical cable faults can be achieved. The deep neural network first preprocesses the OTDR test data from multiple key inspection points in each second-risk optical cable segment by filtering noise and standardizing the data. This transforms the raw reflection signals and attenuation values ​​into a data feature matrix recognizable by the model. The input layer feeds this feature matrix into the hidden layers, where the shallow network extracts basic features such as reflection peak positions and attenuation curve slopes. The deep network further fuses these features to identify higher-order patterns related to the faults. For example, a strong reflection peak combined with steep attenuation may correspond to a breakage fault, while a sustained low attenuation anomaly may correspond to chronic wear. Then, the model uses pre-defined risk assessment indicators to score the risk level of each key inspection point, generating a risk score sequence. Finally, the model can filter out all locations exceeding the risk threshold based on the score sequence and sort them from highest to lowest score. These locations exceeding the risk threshold collectively constitute multiple cable fault points, covering all potential faults requiring priority treatment within the second-risk optical cable segment.

[0104] Based on the same inventive concept Figure 5This is a schematic diagram of an ADSS optical cable inspection system based on data processing, provided in an embodiment of the present invention. The ADSS optical cable inspection system based on data processing includes:

[0105] Acquisition module 91 is used to acquire images along the ADSS optical cable line;

[0106] Risk calculation module 92 is used to determine multiple first-risk optical cable segments and a first fault risk distribution map of each first-risk optical cable segment based on the image along the ADSS optical cable line using a risk distribution model;

[0107] The preliminary detection module 93 is used to determine multiple preliminary detection points for each first-risk optical cable segment based on images of multiple first-risk optical cable segments;

[0108] The first sensing data module 94 is used to acquire distributed fiber strain sensing data of multiple preliminary detection points in each first risk optical cable segment.

[0109] Module 95 is used to construct a first risk map. The first risk map includes multiple preliminary detection nodes and edges between the nodes. The node features of each preliminary detection node are distributed fiber strain sensing data of multiple preliminary detection points of each first risk optical cable segment and ADSS optical cable line images. The edges between the preliminary detection nodes are the relative positional relationships and distances between the preliminary detection nodes.

[0110] The supplementary detection module 96 is used to process the first risk map based on a graph convolutional network to determine multiple supplementary detection points for multiple first-risk optical cable segments, and to acquire distributed optical fiber strain sensing data of multiple supplementary detection points for each first-risk optical cable segment.

[0111] The secondary risk module 97 is used to determine multiple second-risk optical cable segments based on distributed optical fiber strain sensing data from multiple supplementary detection points of each first-risk optical cable segment.

[0112] Repair and positioning module 98 is used to determine the cable fault point based on distributed fiber optic strain sensing data from multiple preliminary detection points of the second risk optical cable segment and distributed fiber optic strain sensing data from multiple supplementary detection points of the second risk optical cable segment.

[0113] It should be noted that, in order to simplify the descriptions disclosed herein and thus aid in the understanding of one or more embodiments of the invention, the foregoing description of embodiments of this specification may sometimes combine multiple features into a single embodiment, drawing, or description thereof. However, this method of disclosure does not imply that the subject matter of this specification requires more features than those mentioned in the claims. In fact, the embodiments contain fewer features than all the features of a single embodiment disclosed above.

[0114] Finally, it should be understood that the embodiments described in this specification are merely illustrative of the principles of the embodiments described herein. Other variations may also fall within the scope of this specification. Therefore, alternative configurations of the embodiments described herein are intended to be illustrative rather than limiting, and should be considered consistent with the teachings of this specification. Accordingly, the embodiments described herein are not limited to those explicitly introduced and described herein.

Claims

1. An ADSS optical cable inspection method based on data processing, characterized in that, include: Acquire images along the ADSS optical cable line; Based on the images along the ADSS optical cable line, a risk distribution model was used to determine multiple first-risk optical cable segments and a first fault risk distribution map for each first-risk optical cable segment. Multiple preliminary detection points for each first-risk optical cable segment are determined based on images of multiple first-risk optical cable segments; Acquire distributed fiber strain sensing data from multiple preliminary detection points in each first-risk fiber optic cable segment; A first risk map is constructed, which includes multiple preliminary detection nodes and edges between the nodes. The node features of each preliminary detection node are distributed fiber strain sensing data and ADSS optical cable line images of multiple preliminary detection points of each first risk optical cable segment. The edges between the preliminary detection nodes are the relative positional relationships and distances between the preliminary detection nodes. The first risk map is processed using a graph convolutional network to determine multiple supplementary detection points for multiple first-risk optical cable segments, and distributed fiber strain sensing data of multiple supplementary detection points for each first-risk optical cable segment are obtained. Multiple second-risk optical cable segments are determined based on distributed optical fiber strain sensing data from multiple supplementary detection points of each first-risk optical cable segment. This determination of multiple second-risk optical cable segments based on distributed optical fiber strain sensing data from multiple supplementary detection points of each first-risk optical cable segment includes: Based on the distributed fiber strain sensing data of multiple preliminary detection points of each first-risk optical cable segment and the distributed fiber strain sensing data of multiple supplementary detection points of each first-risk optical cable segment, a second fault risk distribution map of each first-risk optical cable segment is generated. Based on the first fault risk distribution map of each first-risk optical cable segment and the second fault risk distribution map of each first-risk optical cable segment, multiple second-risk optical cable segments are identified. The cable fault point was determined based on distributed fiber optic strain sensing data from multiple preliminary detection points of the second-risk optical cable segment and distributed fiber optic strain sensing data from multiple supplementary detection points of the second-risk optical cable segment.

2. The ADSS optical cable detection method based on data processing as described in claim 1, characterized in that, The determination of cable fault points based on distributed fiber optic strain sensing data from multiple preliminary detection points of the second-risk optical cable segment and distributed fiber optic strain sensing data from multiple supplementary detection points of the second-risk optical cable segment includes: Acquire environmental videos of multiple secondary risk optical cable segments; Based on the environmental video of the multiple second-risk optical cable segments, the distributed optical fiber strain sensing data of multiple preliminary detection points of the second-risk optical cable segments, and the distributed optical fiber strain sensing data of multiple supplementary detection points of the second-risk optical cable segments, multiple key investigation points for each second-risk optical cable segment are determined. Based on OTDR test data from multiple key inspection points in each second-risk optical cable segment, the cable fault point was determined.

3. The ADSS optical cable detection method based on data processing as described in claim 1, characterized in that, The risk distribution model is a convolutional neural network model.

4. An ADSS optical cable inspection system based on data processing, characterized in that, include: The acquisition module is used to acquire images along the ADSS optical cable line; The risk calculation module is used to determine multiple first-risk optical cable segments and a first fault risk distribution map for each first-risk optical cable segment based on the images along the ADSS optical cable line using a risk distribution model. The preliminary detection module is used to determine multiple preliminary detection points for each first-risk optical cable segment based on images of multiple first-risk optical cable segments; The first sensing data module is used to acquire distributed fiber strain sensing data from multiple preliminary detection points in each first-risk optical cable segment. The construction module is used to construct a first risk map, which includes multiple preliminary detection nodes and edges between the nodes. The node features of each preliminary detection node are distributed fiber strain sensing data and ADSS optical cable line images of multiple preliminary detection points of each first risk optical cable segment. The edges between the preliminary detection nodes are the relative positional relationships and distances between the preliminary detection nodes. The supplementary detection module is used to process the first risk map based on a graph convolutional network to determine multiple supplementary detection points for multiple first-risk optical cable segments, and to acquire distributed optical fiber strain sensing data of multiple supplementary detection points for each first-risk optical cable segment. The secondary risk module is used to determine multiple second-risk optical cable segments based on distributed fiber optic strain sensing data from multiple supplementary detection points in each first-risk optical cable segment. The secondary risk module is also used for: Based on the distributed fiber strain sensing data of multiple preliminary detection points of each first-risk optical cable segment and the distributed fiber strain sensing data of multiple supplementary detection points of each first-risk optical cable segment, a second fault risk distribution map of each first-risk optical cable segment is generated. Based on the first fault risk distribution map of each first-risk optical cable segment and the second fault risk distribution map of each first-risk optical cable segment, multiple second-risk optical cable segments are identified. The repair and positioning module is used to determine the cable fault point based on distributed fiber optic strain sensing data from multiple preliminary detection points of the second-risk optical cable segment and distributed fiber optic strain sensing data from multiple supplementary detection points of the second-risk optical cable segment.

5. The ADSS optical cable inspection system based on data processing as described in claim 4, characterized in that, The repair and positioning module is also used for: Acquire environmental videos of multiple secondary risk optical cable segments; Based on the environmental video of the multiple second-risk optical cable segments, the distributed optical fiber strain sensing data of multiple preliminary detection points of the second-risk optical cable segments, and the distributed optical fiber strain sensing data of multiple supplementary detection points of the second-risk optical cable segments, multiple key investigation points for each second-risk optical cable segment are determined. Based on OTDR test data from multiple key inspection points in each second-risk optical cable segment, the cable fault point was determined.

6. The ADSS optical cable inspection system based on data processing as described in claim 4, characterized in that, The risk distribution model is a convolutional neural network model.

7. An electronic device, characterized in that, include: processor; Memory; And a computer program; wherein the computer program is stored in the memory and configured to be executed by the processor to implement the ADSS optical cable detection method based on data processing as described in any one of claims 1 to 3.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by the processor, the program implements the ADSS optical cable detection method based on data processing as described in any one of claims 1 to 3.

Images