An overhead ground wire metal corrosion residual life prediction method fusing intermittent detection and multi-source timing characteristics

Abstract

This invention discloses a method for predicting the remaining life of overhead ground wire metal corrosion by integrating intermittent detection and multi-source temporal features, belonging to the field of metal corrosion technology. The steps include: acquiring multi-source data of a target overhead ground wire section; acquiring spatial topological information of the section and constructing a graph structure representing the inter-section relationships; expanding the temporal resolution of the intermittent detection data using interpolation methods to match it with continuous monitoring data; then fusing multi-source features on the expanded data to generate aligned temporal fusion features; inputting the temporal fusion features and graph structure into a spatiotemporal prediction model to predict the future corrosion state evolution trend of the section; constructing a health index that meets physical constraints based on this trend; and finally outputting the remaining life prediction result. This invention, through temporal resolution alignment, retains the dynamic identification capability of high-frequency environmental data for the corrosion process, significantly improving the accuracy of predicting the metal corrosion state of overhead ground wires.

Description

Technical Field

[0001] This invention relates to the field of metal corrosion technology, and in particular to a method for predicting the remaining life of overhead ground wire metal corrosion by integrating intermittent detection and multi-source temporal characteristics. Background Technology

[0002] As a crucial protective component of transmission lines, overhead ground wires are exposed to complex outdoor environments for extended periods. Subjected to the combined effects of temperature, humidity, corrosive gases, and rainfall, they are highly susceptible to corrosion of the galvanized layer and rusting of the steel core. This leads to decreased tensile strength and increased risk of wire breakage, seriously threatening the safe and stable operation of transmission lines. Accurately predicting the remaining corrosion life of overhead ground wires is of great significance for ensuring power grid safety and optimizing operation and maintenance strategies.

[0003] In existing technologies for assessing corrosion and predicting the lifespan of overhead ground wires, there is a lack of effective fusion mechanisms between intermittent inspection data and high-frequency environmental time-series data, resulting in low utilization of multi-source data and limited prediction accuracy. Specifically, overhead ground wire corrosion detection is typically conducted on a quarterly or semi-annual basis, and the acquired image-based detection data and corrosion level calibration data are low-frequency intermittent data; while monitoring data such as environmental temperature and humidity and corrosive gas concentrations can be acquired daily or at higher frequencies, belonging to high-frequency time-series data. The temporal resolution of the two types of data differs significantly, and existing technologies often use them separately or only perform simple time aggregation, failing to establish a physical correlation between intermittent detection data and continuous monitoring data. This makes it impossible to fully utilize high-frequency environmental data to identify the corrosion evolution process between two inspections, resulting in corrosion state prediction models that struggle to capture the dynamic impact of environmental fluctuations on the corrosion rate, leading to low accuracy in predicting the corrosion state of overhead ground wires. Summary of the Invention

[0004] To address the problems in existing technologies for predicting the remaining life of overhead ground wire metal corrosion, such as the difficulty in aligning and fusing intermittent detection data with high-frequency continuous monitoring data and the lack of real engineering mechanics basis, resulting in low accuracy in predicting the corrosion state of overhead ground wires, this invention provides a method for predicting the remaining life of overhead ground wire metal corrosion by fusing intermittent detection and multi-source temporal features. This invention achieves multi-modal data fusion through multi-scale feature alignment and an adaptive weighting mechanism, and constructs a physical constraint health index based on the mechanical evolution mechanism of overhead ground wires, significantly improving the accuracy and engineering reliability of the prediction results. The specific technical solution is as follows: This application provides a method for predicting the remaining life of overhead ground wire metal corrosion by integrating intermittent detection and multi-source temporal characteristics, including the following steps: Acquire multi-source data for the target overhead ground wire section. The multi-source data includes intermittent detection data collected in an intermittent manner, continuous monitoring data collected in a continuous manner, and spatial topology information. The intermittent detection data includes at least visible light images, infrared spectral images, calibrated corrosion levels, current tensile strength, and cross-sectional loss rate. The continuous monitoring data includes at least daily average temperature and humidity, corrosive gas concentration, daily rainfall, and sunshine duration. A graph structure representing the inter-segment relationships is constructed based on the spatial topology information of the target overhead ground wire segment; The multi-source data is preprocessed to extract the initial corrosion level and intensity retention rate features from the intermittent detection data, and the continuous monitoring data is standardized. Based on the preprocessed intermittent detection data and the continuous monitoring data, an interpolation method is used to expand the temporal resolution of the intermittent detection data to match the temporal resolution of the continuous monitoring data, and multi-source feature fusion is performed on the expanded data to generate aligned temporal fusion features. The temporal fusion features and the graph structure are input into a pre-trained spatiotemporal prediction model to predict the future corrosion evolution trend of the target overhead ground wire section; wherein, the spatiotemporal prediction model is pre-trained based on a multi-objective joint loss function including physical monotonic constraints; the spatiotemporal prediction model integrates spatial correlation feature extraction and temporal dependency feature extraction; Based on the corrosion state evolution trend, a health index that meets physical constraints is constructed, and the remaining service life prediction result of the target overhead ground wire section is output based on the health index.

[0005] Preferably, the step of using an interpolation method to expand the temporal resolution of the intermittent detection data to match the temporal resolution of the continuous monitoring data, and then performing multi-source feature fusion on the expanded data to generate aligned temporal fusion features, specifically includes: A linear interpolation algorithm is used to expand low-frequency intermittent detection data into a high-frequency pseudo-label sequence, resulting in interpolated corrosion feature sequences and mechanical feature sequences to match the temporal resolution of continuous monitoring data. The interpolated corrosion feature sequence, mechanical feature sequence, and encoded temporal environment sequence are input into the feature fusion unit; The multi-source features are adaptively weighted using an attention mechanism to output a unified temporal fusion feature.

[0006] Preferably, the spatiotemporal prediction model includes a graph neural network module and a time-series autoregressive module; The graph neural network module is used to extract spatial correlation features between different overhead ground wire segments based on the graph structure. The time-series autoregressive module is used to extract time-dependent features based on historical time-series data; the spatiotemporal prediction model outputs the corrosion state prediction results for future moments by fusing the spatial correlation features and the time-dependent features.

[0007] Preferably, a method for predicting the remaining life of overhead ground wire metal corrosion that integrates intermittent detection and multi-source temporal characteristics further includes: Obtain intermittent detection data for the new cycle, and use the intermittent detection data to perform weighted smoothing correction on the current output prediction result of the spatiotemporal prediction model, specifically including: Obtain the actual measured values ​​of corrosion level and reference values ​​of tensile strength from the new round of inspections; A correction weighting coefficient is introduced to linearly weight the predicted result of the current output of the model with the actual measured value of the corrosion level and the reference value of the tensile strength, so as to obtain the updated state value after closed-loop correction, and the updated state value is used as the input of the model in the next time step.

[0008] Preferably, the construction of health indicators that satisfy physical constraints specifically includes: Obtain the current tensile strength, current corrosion level, and cumulative environmental corrosion amount from the previous inspection time to the current time of the target overhead ground wire section; The health index value is calculated based on the weighted combination of the tensile strength term, corrosion grade term, and environmental cumulative corrosion load term. The health index value satisfies the monotonically non-increasing constraint and the value range is normalized. The tensile strength term is determined based on the strength retention rate, the corrosion level term is determined based on the current corrosion level, and the environmental cumulative corrosion load term is determined based on the integral of the environmental corrosion function over the time interval.

[0009] Preferably, the inputs to the environmental corrosion function include one or more of temperature, humidity, rainfall, and corrosive gas concentration, and the output of the environmental corrosion function is a quantitative value used to characterize the contribution of the daily environment to corrosion.

[0010] Preferably, the step of outputting the predicted remaining service life of the target overhead ground wire section based on the health indicators specifically includes: The comprehensive feature vector at the current moment is input into the lifetime prediction layer; The lifespan prediction layer outputs the predicted remaining usable time based on the condition that the health indicator value drops to a preset failure threshold or the strength retention rate drops to a preset safety threshold.

[0011] Preferably, the multi-objective joint loss function specifically includes: The prediction loss term used to measure the prediction bias of health indicators; Life loss term used to measure the bias in remaining life prediction; The monotonically decreasing constraint regularization loss term is used to constrain health indicators to conform to the laws of physical degradation. And the tensile strength loss term used to measure the deviation in tensile strength prediction.

[0012] Preferably, the spatial topology information includes one or more of the following: latitude and longitude, altitude, terrain slope, soil type, distance from industrial pollution source, and micro-topographic features; The micro-topographic features are quantitatively characterized based on the local average slope, local relative elevation difference, and topographic location index within a preset radius around the target section.

[0013] Compared with the prior art, the beneficial effects of the present invention are as follows: To address the temporal resolution differences between low-frequency inspection data and high-frequency environmental data, this invention employs linear interpolation and an attention mechanism for feature alignment and weighted fusion, enabling the model to accurately characterize the profound impact of dynamic environmental fluctuations on the corrosion evolution process. Corrosion level, tensile strength, and cumulative environmental load are uniformly modeled as health indicators, and a monotonically decreasing constraint is introduced, ensuring that lifetime prediction strictly conforms to the objective physical laws of material mechanical degradation, significantly improving the engineering reliability of the prediction results. Simultaneously, this invention introduces a closed-loop iterative correction mechanism based on a new round of intermittent inspection data, establishing a closed-loop iterative mechanism for prediction, measurement, and correction, breaking through the industry bottleneck of error divergence over time. By integrating a multi-objective loss function encompassing health indicator prediction, lifetime prediction, tensile strength, and monotonically constrained parameters, the algorithm learns from the underlying constraint model and follows physical degradation laws, avoiding abnormal prediction outputs that violate engineering common sense. Attached Figure Description

[0014] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the accompanying drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. In all the drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn to scale.

[0015] Figure 1 The flowchart illustrates a method for predicting the remaining life of overhead ground wire metal corrosion by integrating intermittent detection and multi-source temporal features, as provided in an embodiment of the present invention.

[0016] Figure 2 This is a schematic diagram of the overall framework of a method for predicting the remaining life of overhead ground wire metal corrosion that integrates intermittent detection and multi-source temporal features, provided in an embodiment of the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] It should be understood that, when used in this specification, the terms “comprising” and “including” indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0019] It should also be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0020] It should also be further understood that the term "and / or" as used in this specification refers to any combination of one or more of the associated listed items, as well as all possible combinations, and includes such combinations.

[0021] Please refer to the following examples. Figures 1 to 2 .

[0022] This application provides a method for predicting the remaining lifetime of overhead ground wire metal corrosion by integrating intermittent detection and multi-source temporal characteristics, including the following steps: Step S1: Acquire multi-source data for the target overhead ground wire section. The multi-source data includes intermittent detection data collected in an intermittent manner, continuous monitoring data collected in a continuous manner, and spatial topology information. The intermittent detection data includes at least visible light images, infrared spectral images, calibrated corrosion levels, current tensile strength, and cross-sectional loss rate. The continuous monitoring data includes at least daily average temperature and humidity, corrosive gas concentration, daily rainfall, and sunshine duration. In this embodiment, the overhead ground wire section between two adjacent towers is used as the basic object for corrosion status assessment and remaining life prediction, and each quarterly inspection cycle is used as an intermittent inspection sample cycle. For each target section, the intermittent inspection variables, environmental time-series variables, material mechanics variables, spatial geographic variables, and operation and maintenance history variables are acquired and constructed synchronously within the corresponding quarter to form a multi-source input feature set for subsequent corrosion status assessment and remaining life prediction.

[0023] The multi-source variables specifically include: (1) Intermittent inspection variables The intermittent inspection variables are low-frequency detection data collected during each quarterly inspection, including: visible light images, infrared spectral images, corrosion level, current tensile strength, and cross-sectional loss rate. For example, inspection work is carried out on the first day of each quarter, and the data obtained represents the inspection results for that quarter.

[0024] Among them: visible light images are used to extract the area ratio of the rusted area, color distribution characteristics, and the area (diameter / thickness) of surface cracks and eroded overhead ground wires; infrared spectral images are used to extract the corrosion level, area distribution, and area ratio of the rusted area. (2) Environmental time series variables The environmental time-series variables are high-frequency environmental data collected or statistically analyzed daily within the target area, including: daily average temperature, daily average relative humidity, SO2 concentration, and NO. X Concentration, Cl2 concentration, air pollution index, daily rainfall, and daily sunshine duration.

[0025] The aforementioned environmental data was obtained through means such as meteorological stations, environmental monitoring stations, or on-site sensors to characterize the cumulative corrosive environment experienced by the section between two intermittent inspections.

[0026] (3) Material mechanical variables The material mechanical variables include: initial tensile strength, current tensile strength, strength retention rate, and cross-sectional defect rate.

[0027] Among them, the initial tensile strength and yield strength are obtained from the ground wire's factory parameters, design data, or type test data; The current tensile strength is estimated on-site using non-destructive testing methods, or obtained by sampling and destructive testing of failed overhead ground wires on-site. The strength retention rate is defined as: in, Let be the intensity retention rate at time t. The current tensile strength (MPa) This is the initial rated tensile strength (MPa, factory parameter).

[0028] The cross-sectional defect rate is characterized by the proportion of the current effective cross-sectional area to the initial nominal cross-sectional area.

[0029] Currently, tensile strength is estimated by non-destructive testing methods such as ultrasonic, magnetic memory, and electromagnetic testing, combined with a preset strength conversion model, or by directly measuring it through tensile testing equipment by sampling of failed overhead ground wires on site and pre-hanging comparison samples. The cross-sectional loss rate is defined as: in, This is the initial nominal cross-sectional area of ​​the ground wire. This is the current effective cross-sectional area. This refers to the cross-sectional area loss rate.

[0030] (4) Operation and maintenance historical variables The historical operation and maintenance variables include: historical inspection records, corrosion repair records, replacement records, lightning strike records, and icing records.

[0031] Among them, historical inspection records are used to characterize corrosion evolution trends, repair and replacement records are used to characterize local maintenance intervention behaviors, and lightning strike records and icing records are used to characterize the impact of extreme working conditions on the performance of grounding wire structures and corrosion development.

[0032] To ensure the consistency of data from different sources, this invention establishes a unified three-element index of segment number, inspection cycle and timestamp for the above-mentioned multi-source variables, and aggregates the daily environmental time series variables into the corresponding quarterly inspection interval, thereby forming a multi-source time series feature input oriented towards the target segment.

[0033] (5) Spatial topology information The spatial topological information includes: latitude and longitude, altitude, terrain slope, soil type, distance from industrial pollution sources, and micro-topographic features.

[0034] Among them, latitude and longitude are obtained using methods such as Beidou or GPS positioning; altitude is obtained using methods such as digital elevation model (DEM) or field survey data; terrain slope is expressed as the average surface slope within a preset range around the center point of the target section, in degrees; soil type is determined based on soil survey data of the area where the tower base is located below the target section; distance from industrial pollution source is the straight-line distance from the target section to the nearest industrial pollution source, in meters or kilometers; The micro-topographic features are quantitatively characterized by topographic parameters within a preset radius around the center point of the target section. The topographic parameters include at least: local average slope S, local relative elevation difference H, and topographic location index TPI.

[0035] Wherein; the local average slope S is the average surface slope within a radius of 100m around the center point of the target section; the local relative elevation difference H is the difference between the elevation of the highest point and the lowest point within the range; and the topographic position index TPI is the difference between the elevation of the center point of the target section and the average elevation of its surrounding area.

[0036] Based on the combined relationship between TPI and slope S, micro-topographic features are divided into: Valley type: TPI < −5m; Flat type: −5≤TPI≤5m and S<5°; Sloping terrain: −5≤TPI≤5m and S≥5°; Ridge type: TPI>5m.

[0037] The micro-topographic features are used to characterize the effects of water vapor retention, pollutant deposition, and local wind field differences around the target section on the corrosion rate.

[0038] Step S2: Construct a graph structure representing the inter-segment relationships based on the spatial topology information of the target overhead ground wire segment; Specifically, a graph structure G=(V,E) is constructed based on geographic location and route topology to achieve spatial relationship modeling. The node set V={v1, v2, ..., v...} n Each node v in} i Represents an overhead ground wire target segment between adjacent towers; each e in the edge set E ij Represents node v i With node v j There are physical connections or spatial proximity relationships between the target segments. Preferably, when two target segments share the same tower and are adjacent at the beginning and end on the line, a connecting edge is established between the corresponding nodes as the spatial structure input for the subsequent network.

[0039] Step S3: Preprocess the multi-source data, extract the initial corrosion level and intensity retention rate features from the intermittent detection data, and standardize the continuous monitoring data; specifically including: (1) Feature extraction is performed on visible light and infrared spectral images. A dual-branch feature fusion network based on improved YOLOv8 is used to process the visible light and infrared spectral images respectively. The comprehensive corrosion level is output through the feature fusion module based on the material corrosion area, corrosion layer peeling degree, weight loss and diameter change.

[0040] (2) Standardize, interpolate and extract statistical features of environmental time series data using sliding window.

[0041] (3) Construct strength retention rate characteristics from mechanical property data: In the formula, The intensity retention rate at time t is represented, preferably in the form of intensity retention rate. This represents the estimated tensile strength at time t. It is preferred to obtain the results through on-site non-destructive testing.

[0042] Based on the principle of stratified sampling (at least 3-5 points per kilometer, double the number in key areas), priority is given to testing in easily corroded critical areas (such as near wire clamps, vibration zones, and areas where abnormalities have been detected). The remaining wall thickness of the steel core or aluminum wire is measured in relation to the current outer diameter of the conductor to estimate the remaining effective cross-sectional area. Based on the material mechanics relationship that "remaining strength is approximately proportional to remaining cross-sectional area," the current tensile strength is indirectly calculated to obtain an estimated value. ; This indicates the initial rated tensile strength, a factory parameter for overhead ground wires, which can be determined uniformly from the manufacturer's product manual, factory inspection report, or engineering design selection ledger. and Build It is used to characterize the degree of bearing capacity attenuation of overhead ground wires during the corrosion evolution process.

[0043] Step S4: Based on the preprocessed intermittent detection data and the continuous monitoring data, an interpolation method is used to expand the temporal resolution of the intermittent detection data to match the temporal resolution of the continuous monitoring data, and multi-source feature fusion is performed on the expanded data to generate aligned temporal fusion features; A linear interpolation algorithm is used to expand low-frequency intermittent detection data into a high-frequency pseudo-label sequence, resulting in interpolated corrosion feature sequences and mechanical feature sequences to match the temporal resolution of continuous monitoring data. Specifically, to address the temporal resolution differences between quarterly inspection and daily environmental data, linear interpolation and attention-weighted fusion are used to achieve daily feature alignment: In the formula, Indicates the moment when interpolation occurs. Daily corrosion level pseudo-label; Indicates the time of the last inspection. The corresponding actual corrosion level; Indicates the time of this inspection The corresponding actual corrosion level. The aforementioned... and All of these are derived from manually assisted annotations of inspection images, or from calibration values ​​obtained through manual verification of automatic model recognition results; This indicates the time elapsed since the last inspection. The number of days; This represents the total number of days between two consecutive inspections. Using the interpolation method described above, the original quarterly updated corrosion levels can be expanded into a daily pseudo-label sequence, achieving unified alignment with high-frequency environmental time-series data.

[0044] The interpolated corrosion feature sequence, mechanical feature sequence, and encoded temporal environment sequence are input into the feature fusion unit; Multi-source feature fusion is performed through an adaptive weighting mechanism to generate temporal fusion features that include aligned corrosion, mechanical and environmental information. Specifically, attention-weighted fusion is as follows: In the formula, This represents the fused composite feature vector at time t; The time-series environmental characteristics at time t are represented by the output of the environmental variables after being encoded by the LSTM model. This represents the interpolated erosion feature vector at time t, which may contain... Information such as (interpolated corrosion level), corrosion area, corrosion percentage, and corrosion location; The interpolated mechanical characteristics at time t are represented. The corrosion area and corrosion percentage are preferably obtained through a combination of manual calibration and image segmentation results. The corrosion location is obtained by locating the corrosion region in the image and mapping it to the actual spatial location of the overhead ground wire component. The Attention fusion unit is used to... , and Multi-source input features are adaptively weighted to form a unified... .

[0045] Step S5: Input the temporal fusion features and the graph structure into a pre-trained spatiotemporal prediction model to predict the future corrosion state evolution trend of the target overhead ground wire section; wherein, the spatiotemporal prediction model is pre-trained based on a multi-objective joint loss function including physical monotonic constraints; the spatiotemporal prediction model integrates spatial correlation feature extraction and temporal dependency feature extraction; Specifically, the spatiotemporal prediction model adopts a fusion structure of graph convolutional spatial modeling and autoregressive prediction: In the formula, The matrix representing the feature matrix of the model-predicted nodes at time t; The `AR` module represents the true features or features corrected for loop closure at the previous time step; `AR` represents the autoregressive module, used to extract temporal dependencies within the historical window; `L` represents the length of the autoregressive historical window; the LSTM is set with 128 hidden layers and 2 layers to balance feature representation and computational efficiency; `GNN` represents the graph convolutional neural network module, with 8 GAT attention heads, used to model the spatial relationships between overhead ground wire nodes; the training dataset is divided into a 7:2:1 ratio to balance the model's generalization ability with the scarcity of engineering data; training stops when the validation set loss (e.g., MSE, mean squared error) fails to decrease for 10-20 consecutive rounds, reaching the convergence threshold of the loss function, thus avoiding overfitting. By jointly inputting the temporal information extracted by the autoregressive module and the spatial information extracted by the graph convolutional module, spatiotemporal coupled prediction of the corrosion state evolution of overhead ground wires can be achieved.

[0046] Step S6: Based on the corrosion state evolution trend, construct a health index that meets physical constraints, and output the remaining service life prediction result of the target overhead ground wire section based on the health index. Specifically, the construction of health indicators that satisfy physical constraints includes: Obtain the current tensile strength, current corrosion level, and cumulative environmental corrosion amount from the previous inspection time to the current time of the target overhead ground wire section; Currently, tensile strength is estimated using non-destructive testing methods such as ultrasonic testing, magnetic memory testing, and electromagnetic testing, combined with a pre-set strength conversion model. Alternatively, it can be directly measured using tensile testing equipment by sampling failed overhead ground wires in the field and pre-installing comparison samples. Specifically, testing points are set up at key locations in the target section (near the clamp, vibration zone, and areas where abnormalities have been detected) to measure the remaining wall thickness of the steel core or aluminum wire and the current outer diameter of the conductor, thereby estimating the remaining effective cross-sectional area. Based on the material mechanics relationship that residual strength is proportional to residual cross-sectional area, calculate the current tensile strength: in, This represents the remaining tensile strength at the current moment. This is the initial strength at the time of manufacture; Let be the effective remaining cross-sectional area at time t; This refers to the nominal cross-sectional area at the time of manufacture.

[0047] The visible light images and infrared spectral images collected during this inspection are input into a pre-trained dual-branch feature fusion network, and the model outputs a comprehensive corrosion level (0-5).

[0048] From the last inspection time t to the current time t, environmental monitoring data is acquired daily, including daily average temperature, daily average relative humidity, daily average SO2 concentration, and daily average NO. xConcentration and daily rainfall. The daily environmental corrosion contribution is calculated using an environmental corrosion function, and the cumulative environmental corrosion is obtained by summing them up. The specific calculation is as follows: In the formula, For environmental corrosion function (combined factors of temperature, humidity, gas, and rainfall), The output result; Let represent the average daily temperature, relative humidity, corrosive gas concentration, and rainfall on day i, respectively. This is a corrosion rate function for various environmental factors; The weight coefficients of each environmental factor are obtained through model training (such as regression fitting) and, after normalization, satisfy the following conditions: .

[0049] The health index value is calculated based on the weighted combination of the tensile strength term, corrosion grade term, and environmental cumulative corrosion load term. The health index value satisfies the monotonically non-increasing constraint and the value range is normalized. The health index value is calculated by weighting the tensile strength, corrosion grade, and environmental cumulative corrosion load terms: in, Let be a health indicator at time t, with a value range of [0,1]. Tensile strength weight, Corrosion level weighting Environmental cumulative corrosion load weight, satisfying ; Strength retention rate; The current corrosion level; This represents the cumulative corrosion amount since the last inspection.

[0050] By constructing a health index that is a weighted combination of three factors—tensile strength, corrosion level, and environmental cumulative corrosion load—the shortcomings of existing technologies, such as unclear physical meaning of health indicators and inability to comprehensively reflect the coupling effects of multiple factors, are addressed.

[0051] Specifically, based on the health indicators, the remaining service life prediction results of the target overhead ground wire section are output, including: The comprehensive feature vector at the current moment is input into the lifetime prediction layer; The lifespan prediction layer outputs the predicted remaining usable time based on the condition that the health indicator value drops to a preset failure threshold or the strength retention rate drops to a preset safety threshold.

[0052] In practice, when the health indicator value HI drops to the failure threshold or the strength retention rate falls below the safety threshold, the feature vector will be... Input the MLP prediction layer to obtain the remaining lifetime. : ,or In the formula, This represents the current remaining useful life at time t; This indicates the predicted duration of safe operation based on the current state. This indicates the threshold for health indicator failure, which can be set based on historical operating data, engineering safety standards, or empirical thresholds. Indicates the predicted time The strength retention rate is used to determine whether the target overhead ground wire will still meet the requirements for safe operation in the future.

[0053] In a preferred embodiment of the present invention, in order to ensure that the output of the spatiotemporal prediction model strictly conforms to the objective physical laws of corrosion and degradation of overhead ground wire metal materials, and to avoid prediction errors that violate common sense from a purely data-driven model, the spatiotemporal prediction model adopts a multi-objective joint loss function that integrates physical constraints to optimize network parameters during the training phase.

[0054] The specific mathematical expression of the multi-objective joint loss function is as follows: Total losses include losses in health indicators, lifespan, monotonic constraint, and tensile strength. + + + In the formula, Represents the total loss function; This represents the health indicator prediction loss, used to measure the deviation between the model's predicted health indicators and the actual health indicators. This represents the predicted loss of remaining useful life, used to measure the deviation between the predicted life and the actual life. The constraint loss representing the monotonically decreasing health index is a regularization constraint term used to limit the overall non-increasing change of HI during the corrosion evolution process. This represents the tensile strength prediction loss, used to measure the deviation between the tensile strength predicted by the model and the reference true value; , and It can be constructed using mean squared error, mean absolute error, Euclidean distance, or other regression loss forms that characterize the deviation between predicted and true values.

[0055] Step S7: Obtain intermittent detection data for the new cycle, and use the intermittent detection data to perform weighted smoothing correction on the current output prediction state of the spatiotemporal prediction model to achieve closed-loop iterative optimization of the prediction model.

[0056] Specifically, the closed-loop iterative correction mechanism performs weighted correction on the model using actual detection values ​​after each quarterly inspection, achieving iterative optimization that improves accuracy with each test. The weighted correction is as follows: In the formula, This indicates the corrosion level after this closed-loop correction. This indicates the corrosion level predicted by the model. This indicates the actual measured value during inspection or the manually verified calibration value; similarly, This represents the estimated tensile strength after this closed-loop correction. This represents the estimated tensile strength obtained from the model prediction. This indicates the actual measured value during the inspection or the reference value after sampling and calibration. This represents the correction weighting coefficient, which preferably ranges from 0 to 1. It is used to balance the contribution ratio of the model prediction results to the actual inspection values. This can be obtained through manual review or expert annotation of inspection images. It can be obtained from the results of on-site non-destructive testing and can be calibrated and corrected in stages by combining the results of laboratory tensile fracture tests.

[0057] To address the temporal resolution differences between low-frequency inspection data and high-frequency environmental data, this invention employs linear interpolation and an attention mechanism for feature alignment and weighted fusion, enabling the model to accurately characterize the profound impact of dynamic environmental fluctuations on the corrosion evolution process. Corrosion level, tensile strength, and cumulative environmental load are uniformly modeled as health indicators, and a monotonically decreasing constraint is introduced, ensuring that lifetime prediction strictly conforms to the objective physical laws of material mechanical degradation, significantly improving the engineering reliability of the prediction results. A closed-loop iterative correction mechanism based on a new round of intermittent inspection data is introduced, establishing a closed-loop iterative mechanism for prediction, measurement, and correction. By integrating a multi-objective loss function encompassing health indicator prediction, lifetime prediction, tensile strength, and monotonically decreasing constraints, the algorithm learns from the underlying constraint model and follows physical degradation laws, avoiding abnormal prediction outputs that violate engineering common sense and improving the prediction accuracy of overhead ground wire corrosion status.

[0058] Those skilled in the art will recognize that the units of the various examples described in connection with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of the invention.

[0059] In the embodiments provided by the present invention, it should be understood that the division of units is only a logical functional division. In actual implementation, there may be other division methods, such as multiple units can be combined into one unit, one unit can be split into multiple units, or some features can be ignored.

[0060] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0061] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0062] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be covered within the scope of the specification of the present invention.

Claims

1. A method for predicting the remaining life of overhead ground wire metal corrosion by integrating intermittent detection and multi-source temporal characteristics, characterized in that, Includes the following steps: Acquire multi-source data for the target overhead ground wire section. The multi-source data includes intermittent detection data collected in an intermittent manner, continuous monitoring data collected in a continuous manner, and spatial topology information. The intermittent detection data includes at least visible light images, infrared spectral images, calibrated corrosion levels, current tensile strength, and cross-sectional loss rate. The continuous monitoring data includes at least daily average temperature and humidity, corrosive gas concentration, daily rainfall, and sunshine duration. A graph structure representing the inter-segment relationships is constructed based on the spatial topology information of the target overhead ground wire segment; The multi-source data is preprocessed to extract the initial corrosion level and intensity retention rate features from the intermittent detection data, and the continuous monitoring data is standardized. Based on the preprocessed intermittent detection data and the continuous monitoring data, an interpolation method is used to expand the temporal resolution of the intermittent detection data to match the temporal resolution of the continuous monitoring data, and multi-source feature fusion is performed on the expanded data to generate aligned temporal fusion features. The temporal fusion features and the graph structure are input into a pre-trained spatiotemporal prediction model to predict the future corrosion evolution trend of the target overhead ground wire section; wherein, the spatiotemporal prediction model is pre-trained based on a multi-objective joint loss function including physical monotonic constraints; the spatiotemporal prediction model integrates spatial correlation feature extraction and temporal dependency feature extraction; Based on the corrosion state evolution trend, a health index that meets physical constraints is constructed, and the remaining service life prediction result of the target overhead ground wire section is output based on the health index.

2. The method for predicting the remaining lifetime of overhead ground wire metal corrosion by integrating intermittent detection and multi-source temporal characteristics according to claim 1, characterized in that, The step of using an interpolation method to extend the temporal resolution of the intermittent detection data to match the temporal resolution of the continuous monitoring data, and then performing multi-source feature fusion on the extended data to generate aligned temporal fusion features, specifically includes: A linear interpolation algorithm is used to expand low-frequency intermittent detection data into a high-frequency pseudo-label sequence, resulting in interpolated corrosion feature sequences and mechanical feature sequences to match the temporal resolution of continuous monitoring data. The interpolated corrosion feature sequence, mechanical feature sequence, and encoded temporal environment sequence are input into the feature fusion unit; The multi-source features are adaptively weighted using an attention mechanism to output a unified temporal fusion feature.

3. The method for predicting the remaining lifetime of overhead ground wire metal corrosion by integrating intermittent detection and multi-source temporal characteristics according to claim 1, characterized in that, The spatiotemporal prediction model includes a graph neural network module and a time-series autoregressive module; The graph neural network module is used to extract spatial correlation features between different overhead ground wire segments based on the graph structure. The time-series autoregressive module is used to extract time-dependent features based on historical time-series data; the spatiotemporal prediction model outputs the corrosion state prediction results for future moments by fusing the spatial correlation features and the time-dependent features.

4. The method for predicting the remaining lifetime of overhead ground wire metal corrosion by integrating intermittent detection and multi-source temporal characteristics according to claim 1, characterized in that, Also includes: Obtain intermittent detection data for the new cycle, and use the intermittent detection data to perform weighted smoothing correction on the current output prediction result of the spatiotemporal prediction model, specifically including: Obtain the actual measured values ​​of corrosion level and reference values ​​of tensile strength from the new round of inspections; A correction weighting coefficient is introduced to linearly weight the predicted result of the current output of the model with the actual measured value of the corrosion level and the reference value of the tensile strength, so as to obtain the updated state value after closed-loop correction, and the updated state value is used as the input of the model in the next time step.

5. The method for predicting the remaining lifetime of overhead ground wire metal corrosion by integrating intermittent detection and multi-source temporal characteristics according to claim 1, characterized in that, The construction of health indicators that satisfy physical constraints specifically includes: Obtain the current tensile strength, current corrosion level, and cumulative environmental corrosion amount from the previous inspection time to the current time of the target overhead ground wire section; The health index value is calculated based on the weighted combination of the tensile strength term, corrosion grade term, and environmental cumulative corrosion load term. The health index value satisfies the monotonically non-increasing constraint and the value range is normalized. The tensile strength term is determined based on the strength retention rate, the corrosion level term is determined based on the current corrosion level, and the environmental cumulative corrosion load term is determined based on the integral of the environmental corrosion function over the time interval.

6. The method for predicting the remaining lifetime of overhead ground wire metal corrosion by integrating intermittent detection and multi-source temporal characteristics according to claim 5, is characterized in that, The inputs to the environmental corrosion function include one or more of temperature, humidity, rainfall, and corrosive gas concentration, and the output of the environmental corrosion function is a quantitative value used to characterize the degree of contribution of the environment to corrosion on a single day.

7. The method for predicting the remaining life of overhead ground wire metal corrosion by integrating intermittent detection and multi-source temporal characteristics according to claim 5, characterized in that, The step of outputting the remaining service life prediction result of the target overhead ground wire section based on the health indicators specifically includes: The comprehensive feature vector at the current moment is input into the lifetime prediction layer; The lifespan prediction layer outputs the predicted remaining usable time based on the condition that the health indicator value drops to a preset failure threshold or the strength retention rate drops to a preset safety threshold.

8. The method for predicting the remaining lifetime of overhead ground wire metal corrosion by integrating intermittent detection and multi-source temporal characteristics according to claim 1, characterized in that, The multi-objective joint loss function specifically includes: The prediction loss term used to measure the prediction bias of health indicators; Life loss term used to measure the bias in remaining life prediction; The monotonically decreasing constraint regularization loss term is used to constrain health indicators to conform to the laws of physical degradation. And the tensile strength loss term used to measure the deviation in tensile strength prediction.

9. The method for predicting the remaining lifetime of overhead ground wire metal corrosion by integrating intermittent detection and multi-source temporal characteristics according to claim 1, characterized in that, The spatial topology information includes one or more of the following: latitude and longitude, altitude, terrain slope, soil type, distance from industrial pollution sources, and micro-topographic features; The micro-topographic features are quantitatively characterized based on the local average slope, local relative elevation difference, and topographic location index within a preset radius around the target section.

Images