Cement-based telegraph pole defect detection method based on big data analysis

Abstract

The invention discloses a cement-based telegraph pole defect detection method based on big data analysis, and the method comprises the steps: arranging a distributed optical fiber sensing network to collect the stress data of a target telegraph pole, and obtaining a stress feature set through denoising and feature extraction; constructing a structure damage feature library based on the attributes of the telegraph pole and the material characteristics, and screening key damage features through a self-adaptive matching threshold value; and in combination with concrete carbonization characteristics and steel bar corrosion characteristics, carrying out weighted analysis by adopting a multi-modal fusion model and an attention mechanism, and outputting an optimally matched defect detection result. According to the method, through multi-source data fusion and dynamic threshold optimization, the recognition precision of typical damages such as cracks and spalling is remarkably improved, a graded early warning and maintenance scheme can be automatically generated according to a detection result, and intelligent monitoring and operation and maintenance of the health state of the telegraph pole are achieved; the method has the characteristics of high detection accuracy and strong environmental adaptability, and is suitable for long-term safety monitoring of the electric power infrastructure.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of cement-based power pole defect detection, in particular to a cement-based power pole defect detection method based on big data analysis. BACKGROUND

[0002] With the rapid development of power infrastructure, cement-based power poles as an important part of the power transmission network, the structural safety is directly related to the reliable operation of the power grid. However, long-term exposure to complex environments, power poles are prone to concrete carbonation, steel corrosion, crack propagation and other damages, and serious accidents such as pole body rupture may occur. The traditional detection method mainly relies on manual inspection and local detection technology, which has the problems of low efficiency, strong subjectivity and difficulty in finding early hidden defects.

[0003] At present, the structure health monitoring method based on sensing technology has been gradually applied to the field of infrastructure detection. In the prior art, some schemes use a single type of sensor (such as a vibration sensor or a strain gauge) to collect data, but this method can only reflect the damage characteristics of a specific type, and it is difficult to comprehensively evaluate the overall damage state of the power pole. In addition, the conventional data analysis method usually uses a fixed threshold for damage judgment, which cannot adapt to the data fluctuations caused by different environmental conditions and material properties, resulting in high false alarm rate and low detection rate.

[0004] In terms of data processing, the existing technology is limited to the analysis of a single physical quantity, and lacks effective fusion of multi-source heterogeneous data (such as stress, carbonation depth, corrosion rate, etc.). Although a few studies have tried to combine multiple sensor data for damage assessment, a scientific feature weight distribution mechanism has not been established, making it difficult to accurately quantify the contribution of different damage characteristics. At the same time, the utilization rate of historical data by traditional methods is low, and the deep correlation between sample characteristics and damage patterns accumulated in the big data environment has not been fully tapped. In addition, the existing technology has the problems of low defect detection and recognition accuracy, low defect detection accuracy, and poor environmental adaptability, especially the inability to quickly and accurately detect hidden defects of aging power poles. SUMMARY

[0005] The purpose of the present application is to provide a cement-based power pole defect detection method based on big data analysis to solve the above problems existing in the prior art.

[0006] The present application is as follows:

[0007] A cement-based power pole defect detection method based on big data analysis, the method comprising:

[0008] A distributed sensing network is arranged based on predetermined monitoring points, and the target electric pole is continuously monitored through the distributed sensing network to obtain a plurality of stress data sets within a preset period;

[0009] Stress feature extraction is performed on the plurality of stress data sets to obtain a plurality of stress feature sets;

[0010] The plurality of stress feature sets are respectively input into a structural damage feature library for multi-dimensional feature matching, a plurality of damage feature sets that satisfy an adaptive matching threshold are selected, and cross-set analysis is performed on the plurality of damage feature sets to output a plurality of key damage features;

[0011] The concrete carbonation features and steel bar corrosion features of the target electric pole within a preset period are collected, and multi-modal fusion analysis is performed on the plurality of key damage features according to the concrete carbonation features and steel bar corrosion features to output an optimally matched key damage feature as a cement-based electric pole defect detection result of the target electric pole.

[0012] Further, a distributed sensing network is arranged based on predetermined monitoring points, and the target electric pole is continuously monitored through the distributed sensing network to obtain a plurality of stress data sets within a preset period, including:

[0013] A predetermined monitoring point of the target electric pole is obtained, wherein the predetermined monitoring point at least includes a middle part of the pole body, a bottom part of the pole body, and a pole body connection part;

[0014] Optical fiber stress sensors are arranged at the middle part of the pole body, the bottom part of the pole body, and the pole body connection part to construct a distributed sensing network;

[0015] The target electric pole is continuously monitored through the distributed sensing network at a predetermined monitoring frequency to obtain a plurality of stress data sets of a plurality of monitoring points within a preset period.

[0016] Further, stress feature extraction is performed on the plurality of stress data sets to obtain a plurality of stress feature sets, including:

[0017] Signal interference analysis and filtering denoising processing are performed on the plurality of stress data sets to obtain a plurality of signal interference ratios and a plurality of standard stress data sets;

[0018] A stress feature index is configured, wherein the stress feature index at least includes an abnormal stress component, an abnormal stress proportion, a stress change curve, and a waveform curve;

[0019] According to the stress feature index, stress feature extraction is performed on the plurality of standard stress data sets to obtain a plurality of stress feature sets of the plurality of monitoring points.

[0020] Further, the plurality of stress feature sets are respectively input into a structure damage feature library for multi-dimensional feature matching, and a plurality of damage feature sets satisfying an adaptive matching threshold are selected, including:

[0021] A structure damage feature library is constructed based on the power pole attribute features, cement-based material attribute features, and the stress feature indicators;

[0022] The plurality of signal interferences are subjected to mean value calculation to determine a comprehensive signal-to-noise ratio, and a ratio of the comprehensive signal-to-noise ratio to a standard signal-to-noise ratio is calculated to obtain a threshold compensation coefficient;

[0023] The initial similarity comparison threshold is compensated according to the threshold compensation coefficient to obtain an adaptive matching threshold;

[0024] The plurality of stress feature sets are respectively input into the structure damage feature library for multi-dimensional feature matching, and a damage feature satisfying the adaptive matching threshold in terms of matching degree is selected, and the plurality of damage feature sets are output, wherein the damage feature includes a damage type and a damage degree of the cement-based power pole, and the damage type at least includes a crack, a spalling, a deformation, and a corrosion.

[0025] Further, a structure damage feature library is constructed based on the power pole attribute features, cement-based material attribute features, and the stress feature indicators, including:

[0026] The power pole attribute features and the cement-based material attribute features are taken as attribute constraints, the stress feature indicators are taken as conditional constraints, cement-based power pole damage detection based on stress analysis is taken as a guide, information retrieval is performed based on big data, a plurality of sample stress feature sets are obtained, and damage types and damage degrees of cement-based power poles under different sample stress feature sets are labeled to obtain a plurality of sample damage types and a plurality of sample damage degrees;

[0027] The plurality of sample damage types and the plurality of sample damage degrees are clustered based on the same damage type and damage degree interval to determine a plurality of standard damage types, wherein each standard damage type is identified with a plurality of damage degree intervals;

[0028] A first standard damage type and a first damage degree interval of the first standard damage type are randomly selected, the plurality of sample stress feature sets are filtered according to the first standard damage type and the first damage degree interval to determine a plurality of first sample stress feature sets;

[0029] The plurality of first sample stress feature sets are subjected to high-frequency feature analysis to determine a first standard stress feature set, wherein the first standard stress feature set includes a first standard abnormal stress component, a first standard abnormal stress proportion, a first standard stress change curve, and a first standard waveform curve;

[0030] establishing a first mapping association between the first standard stress feature set and the first standard damage type and the first damage degree interval, and constructing a first damage identification branch according to the first mapping association;

[0031] sequentially analyzing a plurality of damage identification branches, and constructing the structural damage feature library according to the plurality of damage identification branches.

[0032] Further, the plurality of stress feature sets are respectively input into the structural damage feature library for multi-dimensional feature matching, and a damage feature satisfying the adaptive matching threshold is selected, and the plurality of damage feature sets are output, including:

[0033] randomly selecting a first stress feature set, and randomly selecting a first damage identification branch in the structural damage feature library;

[0034] inputting the first stress feature set into the first damage identification branch, and performing multi-dimensional feature matching with the first standard stress feature set, if the matching degrees all satisfy the adaptive matching threshold, then setting the first standard damage type and the first damage degree interval as a first damage feature;

[0035] sequentially performing multi-dimensional feature matching between the first stress feature set and the standard stress feature sets of other damage identification branches in the structural damage feature library, obtaining a first damage feature set, and adding the first damage feature set to the plurality of damage feature sets.

[0036] Further, the first stress feature set is input into the first damage identification branch, and multi-dimensional feature matching is performed with the first standard stress feature set, if the matching degrees all satisfy the adaptive matching threshold, then the first standard damage type and the first damage degree interval are set as a first damage feature, including:

[0037] obtaining the first stress feature set, wherein the first stress feature set includes a first abnormal stress component, a first abnormal stress proportion, a first stress change curve, and a first waveform curve;

[0038] performing feature matching on the first abnormal stress component and a first standard abnormal stress component to determine a first matching degree;

[0039] if the first matching degree is greater than the adaptive matching threshold, then performing feature matching on the first abnormal stress proportion and a first standard abnormal stress proportion to determine a second matching degree;

[0040] if the second matching degree is greater than the adaptive matching threshold, then performing feature matching on the first stress change curve and a first standard stress change curve to determine a third matching degree;

[0041] If the third matching degree is greater than the adaptive matching threshold, feature matching is performed on the first waveform curve and the first standard waveform curve to determine a fourth matching degree, and if the fourth matching degree is greater than the adaptive matching threshold, the first standard damage type and the first damage degree interval are set as the first damage feature.

[0042] Further, the plurality of key damage features are subjected to multi-modal fusion analysis according to the concrete carbonation features and the reinforcement corrosion features, and the optimally matched key damage feature is output as the cement-based utility pole defect detection result of the target utility pole, including:

[0043] The sample concrete carbonation feature set, the sample reinforcement corrosion feature set, and the plurality of sample damage feature sets are retrieved and obtained by taking the utility pole attribute features and the cement-based material attribute features as attribute constraints, taking the damage type and the damage degree as feature constraints, and taking the cement-based utility pole damage detection as guidance;

[0044] A multi-modal feature fusion model is constructed, and the stress features, the concrete carbonation features, and the reinforcement corrosion features are taken as input modalities;

[0045] The attention mechanism is adopted to calculate the weight coefficients of the modal features;

[0046] The modal features are weighted and fused based on the weight coefficients to calculate the matching degree scores of the key damage features;

[0047] The key damage feature with the highest matching degree score is selected as the optimal matching result, which is set as the cement-based utility pole defect detection result of the target utility pole.

[0048] Further, the method further includes:

[0049] An damage warning level is generated based on the cement-based utility pole defect detection result;

[0050] A corresponding maintenance scheme in a maintenance strategy library is matched according to the damage warning level;

[0051] The maintenance scheme is associated with the utility pole location information and then pushed to a maintenance terminal.

[0052] Further, the damage warning level is generated based on the cement-based utility pole defect detection result, including:

[0053] A mapping relationship table of the damage type, the damage degree, and the warning level is established;

[0054] The mapping relationship table is queried according to the damage type and the damage degree in the cement-based utility pole defect detection result;

[0055] The warning level is corrected in combination with environmental factors of the utility pole, and the final damage warning level is output.

[0056] Compared with the prior art, the present application has the following beneficial effects:

[0057] The present application arranges a distributed optical fiber sensing network to collect stress data of the target power pole, and obtains a stress feature set through denoising and feature extraction. A structural damage feature library is constructed based on the properties of the power pole and the material characteristics, and key damage features are screened through adaptive matching threshold. Combined with the concrete carbonation features and the steel corrosion features, a multi-modal fusion model and an attention mechanism are used for weighted analysis to output the optimal matching defect detection result. Through multi-source data fusion and dynamic threshold optimization, the present application significantly improves the recognition accuracy of typical damages such as cracks and peeling, and can automatically generate a grading warning and maintenance scheme according to the detection result, realizing intelligent monitoring and operation of the health status of the power pole. The method has the characteristics of high detection accuracy and strong environmental adaptability, and is suitable for long-term safety monitoring of power infrastructure. BRIEF DESCRIPTION OF DRAWINGS

[0058] Figure 1 is a flowchart of a cement-based power pole defect detection method based on big data analysis provided by an embodiment of the present application. DETAILED DESCRIPTION

[0059] The present application will be described in detail below with reference to the accompanying drawings.

[0060] Embodiment 1

[0061] An embodiment of the present application provides a cement-based power pole defect detection method based on big data analysis, as shown in Figure 1 The method comprises the following steps:

[0062] S100: A distributed sensing network is arranged based on a predetermined monitoring point, and the target power pole is continuously monitored through the distributed sensing network to obtain a plurality of stress data sets within a preset period.

[0063] Further, step S100 of the present application further comprises:

[0064] S110: A predetermined monitoring point of the target power pole is obtained, wherein the predetermined monitoring point at least includes a middle part of the pole body, a bottom part of the pole body and a connection part of the pole body;

[0065] S120: Fiber optic stress sensors are arranged at the middle part of the pole body, the bottom part of the pole body and the connection part of the pole body respectively to construct a distributed sensing network;

[0066] S130: The target power pole is continuously monitored through the distributed sensing network according to a predetermined monitoring frequency to obtain a plurality of stress data sets of a plurality of monitoring points within a preset period.

[0067] Specifically, first, the predetermined monitoring points of the target power pole are acquired, wherein the predetermined monitoring points at least include the middle part of the pole body, the bottom of the pole body and the pole body connection, which cover the key stress positions of the power pole and can effectively reflect the overall structural state of the power pole and the related signals of potential defects. Then, optical fiber stress sensors are arranged at the middle part of the pole body, the bottom of the pole body and the pole body connection respectively, and each sensor is connected to a data acquisition system to build a comprehensive distributed sensing network. The network coordinates the work of each sensor to ensure that the stress signals of each key position can be collected and analyzed in real time. Then, according to the structural characteristics of the power pole, a suitable monitoring frequency (such as collecting a certain amount of stress data per minute) is set to ensure that the stress changes of the power pole under various environmental conditions, especially the abnormal stress signals related to structural defects, can be captured. Finally, according to the predetermined monitoring frequency, the target power pole is continuously monitored through the distributed sensing network to obtain multiple stress data sets of multiple monitoring points (the middle part of the pole body, the bottom of the pole body and the pole body connection) within a predetermined period (such as the next 24 hours). By obtaining multiple stress data sets of multiple monitoring points, basic data support is provided for subsequent defect analysis.

[0068] S200: stress feature extraction is performed on the multiple stress data sets to obtain multiple stress feature sets.

[0069] Further, step S200 of the present application further comprises:

[0070] S210: signal interference analysis and filtering denoising processing are performed on the multiple stress data sets to obtain multiple signal interference ratios and multiple standard stress data sets.

[0071] S220: stress feature indexes are configured, wherein the stress feature indexes at least include abnormal stress components, abnormal stress proportions, stress change curves and waveform curves.

[0072] S230: stress feature extraction is performed on the multiple standard stress data sets according to the stress feature indexes to obtain multiple stress feature sets of the multiple monitoring points.

[0073] Specifically, first, signal interference analysis and filtering denoising processing are performed on the plurality of stress data sets respectively, such as using statistical analysis methods (such as mean, standard deviation analysis) to identify abnormal fluctuations or signal interference in the data, common signal interference includes sensor failure, abnormal signals caused by environmental interference, etc., the higher the signal interference ratio, the more signal interference components in the signal; appropriate filtering algorithm is used to filter the stress data, remove high-frequency signal interference or low-frequency interference, common filtering methods include wavelet transform, Kalman filtering, mean filtering, etc.; a plurality of signal interference ratios and a plurality of standard stress data sets are obtained, wherein the signal interference ratio reflects the relative influence degree of signal interference in the data, and the standard stress data set is a clear signal after denoising processing.

[0074] Then, configure stress feature indicators, wherein the stress feature indicators are used for subsequent feature extraction and defect analysis, these feature indicators can effectively capture abnormal changes and potential defect signs in the stress signal, at least including abnormal stress component, abnormal stress proportion, stress change curve and waveform curve, the abnormal stress component is usually related to the structure defect of the power pole (such as crack, peeling, etc.), the frequency component of the stress signal is extracted by Fourier transform or fast Fourier transform (FFT), the specific abnormal stress component is identified and compared with the stress mode of the normal power pole; the abnormal stress proportion is the proportion of the abnormal stress component in the whole frequency spectrum, which is used to indicate the severity of the abnormality, the higher the proportion, the greater the possibility of defect; the stress change curve can intuitively show the stress intensity change of the power pole in different time periods, abnormal stress change is often related to structure defect or imbalance; the waveform curve reflects the change of the stress signal with time, by analyzing the shape of the waveform curve, the defect mode of the power pole structure can be identified.

[0075] Then, according to the stress feature indicators, stress feature extraction is performed on the plurality of standard stress data sets, such as using Fourier transform (FFT) method to extract the frequency component of the stress signal, identifying the abnormal stress component and the corresponding amplitude; the stress signal change is analyzed by time domain statistical features (such as peak value, root mean square value, standard deviation, etc.), to identify possible stress abnormalities; the shape of the stress signal waveform is analyzed to detect whether there is irregular stress waveform; then, combined with time domain, frequency domain and waveform analysis, multi-dimensional stress features of each monitoring point (middle of the pole body, bottom of the pole body and connection of the pole body) are extracted, to obtain a plurality of stress feature sets of a plurality of monitoring points, wherein the stress feature set includes abnormal stress component, abnormal stress proportion, stress change curve and waveform curve, etc., these feature sets provide detailed basis for subsequent defect identification and diagnosis.

[0076] S300: input the plurality of stress feature sets into a structural damage feature library respectively for multi-dimensional feature matching, select a plurality of damage feature sets satisfying an adaptive matching threshold, and perform cross-set analysis on the plurality of damage feature sets, and output a plurality of key damage features.

[0077] It should be noted that the specific technical steps of the cross-set analysis are as follows:

[0078] A1: performing a union operation on the plurality of damage feature sets (such as feature sets of different damage types such as cracks and peeling) matched from the structural damage feature library to form a candidate feature pool; aligning and normalizing the dimensions such as damage type, damage degree, and spatial position in each key damage feature set;

[0079] A2: checking the logical consistency of damage features output by different sensor monitoring points (such as the middle and bottom of the rod) in time and space;

[0080] Example: If corrosion features are detected at the bottom of the rod, and high stress deformation features are detected at the middle of the rod, it is necessary to verify whether they can coexist by combining the material mechanics model;

[0081] Probability weighted voting: for different candidate features of the same damage type, weighted voting is performed according to the matching degree score, and the feature with the highest confidence is selected;

[0082] A3: taking stress features (such as abnormal stress waveforms), concrete carbonation features (such as carbonation depth), and steel corrosion features (such as corrosion rate) as independent evidence chains, and performing correlation analysis in the following ways:

[0083] Joint probability model: calculate the joint probability distribution of each key damage feature under multiple modalities, and eliminate low probability combinations;

[0084] Decision tree rule: preset rule library (such as "carbonation depth > 5mm and stress anomaly → preferentially determine as peeling"), and perform logical reasoning on conflicting features;

[0085] A4: secondary screening of the matching degree of candidate features: if a damage feature exceeds the threshold in multiple monitoring points or modalities, its priority is increased; if the feature is matched in a single modality, the reliability of the feature is judged in combination with environmental data (such as temperature and humidity);

[0086] A5: finally outputting a plurality of key damage features after cross-validation, for subsequent multi-modal fusion analysis, to ensure that the results have spatial coverage (supported by different monitoring points) and modal consistency (stress, carbonation, and corrosion features are mutually verified).

[0087] Further, the step S300 of the present application further comprises:

[0088] S310: Construct a structure damage feature library based on the electric pole attribute features, the cement-based material attribute features, and the stress feature indexes.

[0089] Further, step S310 of the present application further includes:

[0090] S311: Based on the electric pole attribute features and the cement-based material attribute features as attribute constraints, based on the stress feature indexes as condition constraints, based on cement-based electric pole damage detection based on stress analysis as guidance, based on big data to perform information retrieval, obtain a plurality of sample stress feature sets, and label the damage types and damage degrees of the cement-based electric poles under different sample stress feature sets to obtain a plurality of sample damage types and a plurality of sample damage degrees;

[0091] S312: Cluster the plurality of sample damage types and the plurality of sample damage degrees based on the same damage type and damage degree interval, and determine a plurality of standard damage types, wherein each standard damage type is identified with a plurality of damage degree intervals;

[0092] S313: Randomly select a first standard damage type and a first damage degree interval of the first standard damage type, filter the plurality of sample stress feature sets according to the first standard damage type and the first damage degree interval, and determine a plurality of first sample stress feature sets;

[0093] S314: Perform high-frequency feature analysis on the plurality of first sample stress feature sets to determine a first standard stress feature set, wherein the first standard stress feature set includes a first standard abnormal stress component, a first standard abnormal stress proportion, a first standard stress change curve, and a first standard waveform curve;

[0094] S315: Establish a first mapping association between the first standard stress feature set and the first standard damage type and the first damage degree interval, and construct a first damage recognition branch according to the first mapping association;

[0095] S316: Analyze a plurality of damage recognition branches in sequence, and construct the structure damage feature library according to the plurality of damage recognition branches.

[0096] Specifically, first, the electric pole attribute features and the cement-based material attribute features are acquired, wherein the electric pole attribute features include electric pole type, service life, structural characteristics, etc.; the cement-based material attribute features include concrete strength, reinforcement configuration, etc. Then, the electric pole attribute features and the cement-based material attribute features are taken as attribute constraints, i.e. the collection range and type of stress features are limited through device constraints, the stress feature index is taken as a condition constraint, and based on the stress analysis of the cement-based electric pole damage detection, information retrieval is performed based on big data, i.e. based on the electric pole service state, the monitoring data of the stress sensor, and the different working conditions of the electric pole, the stress data sets of multiple electric poles are collected through a data acquisition system; and after data cleaning and preprocessing are performed by using a big data analysis platform, the stress features meeting the device constraints and the condition constraints are extracted to form multiple sample stress feature sets. Further, the damage types and damage degrees of the cement-based electric poles under different sample stress feature sets are labeled, i.e. each stress feature set is manually labeled or classified through an intelligent analysis algorithm (such as a machine learning algorithm), and the damage types (such as cracks, peeling, deformation, corrosion, etc.) and the corresponding damage degrees (mild, moderate, severe) of the cement-based electric poles are labeled; multiple sample damage types and multiple sample damage degrees are acquired.

[0097] Then, the multiple sample damage types and the multiple sample damage degrees are clustered based on the same damage type and damage degree interval, such as using a clustering algorithm (such as K-means, K-Medoids, OPTICS, hierarchical clustering, etc.) to perform clustering analysis on the damage types and degrees. The purpose of this step is to classify samples with similar features into the same class, so as to identify multiple standard damage types and the corresponding damage degree intervals. The clustering analysis maps the relationship between the damage types and the damage degrees, ensures that the same type of damage is gathered together, and according to the change of the damage degree, different degree intervals are divided to determine multiple standard damage types (such as cracks, peeling, deformation, corrosion), wherein each standard damage type is identified with multiple damage degree intervals.

[0098] Then any one of the plurality of standard damage types is randomly selected as a first standard damage type, and a first damage degree interval (any one degree interval) of the first standard damage type is randomly selected; then the plurality of sample stress feature sets are screened according to the first standard damage type and the first damage degree interval to determine a plurality of first sample stress feature sets. Further, high-frequency feature analysis is performed on the plurality of first sample stress feature sets, and through the high-frequency feature analysis on the plurality of first sample stress feature sets, high-frequency components related to damage are extracted, and these frequency components can generally reflect the nature and degree of damage, such as selecting a mode value in the plurality of first sample stress feature sets as a first standard stress feature set to obtain the first standard stress feature set, wherein the first standard stress feature set includes a first standard abnormal stress component, a first standard abnormal stress proportion, a first standard stress change curve, and a first standard waveform curve.

[0099] Further, a first mapping association between the first standard stress feature set and the first standard damage type and the first damage degree interval is established, the mapping relationship associates different stress features with specific types and degrees of damage, and provides a basis for subsequent damage diagnosis. Then, based on the decision tree principle, the first standard stress feature set is taken as a child node, the first standard damage type and the first damage degree interval are taken as leaf nodes of the child node, a first damage recognition branch is constructed according to the first mapping association, the branch is a conditional constraint based on the first standard stress feature set and the damage type and degree, and can accurately recognize damage for a specific damage type and degree. A plurality of damage recognition branches are sequentially analyzed and constructed by using the same method, and then the plurality of damage recognition branches are integrated into a complete structural damage feature library, the library contains different damage types, different damage degrees, and corresponding stress features, thereby providing a comprehensive basis for online defect detection of the cement-based power pole.

[0100] S320: The mean value of the plurality of signal interferences is calculated to determine a comprehensive signal-to-noise ratio, the ratio of the comprehensive signal-to-noise ratio to a standard signal-to-noise ratio is calculated to obtain a threshold compensation coefficient;

[0101] S330: The initial similarity comparison threshold is compensated according to the threshold compensation coefficient to obtain an adaptive matching threshold.

[0102] Specifically, in the stress data, signal interference may affect the extraction of stress features and the accuracy of damage feature recognition, so it is necessary to analyze the signal interference and calculate the comprehensive signal-to-noise ratio to evaluate the influence of signal interference and compensate in subsequent processing. First, the mean value of the plurality of signal interference ratios is calculated, and the mean value calculation result is set as the comprehensive signal-to-noise ratio, then the ratio of the comprehensive signal-to-noise ratio and the standard signal-to-noise ratio (ideal signal interference level obtained based on historical data or theoretical analysis) is calculated to obtain a threshold compensation coefficient, which reflects the difference between the actual signal interference and the standard signal interference, and can be used to adjust the matching threshold. Further, the initial similarity comparison threshold is compensated according to the threshold compensation coefficient, that is, the initial similarity comparison threshold is multiplied by the reciprocal of the threshold compensation coefficient, for example, assuming that the threshold compensation coefficient is 0.95 and the initial similarity comparison threshold is 80%, which represents high data quality, so the similarity comparison threshold needs to be improved, then the reciprocal of 0.95 is multiplied by 80% to obtain about 0.84%, which is set as the adaptive matching threshold; the adaptive matching threshold is obtained, wherein the larger the reciprocal of the threshold compensation coefficient, the higher the adaptive matching threshold, which makes the damage recognition more strict and reduces the possibility of false positives; the smaller the reciprocal of the threshold compensation coefficient, the lower the adaptive matching threshold, which makes the tolerance of damage recognition higher, thereby reducing the false negatives. Through this adaptive matching threshold adjustment method, the signal interference changes in the pole monitoring can be effectively dealt with, the flexibility and accuracy of defect detection are improved, and in a complex signal interference environment, the adaptive matching threshold is automatically adjusted to ensure the accuracy of the system and avoid missing detection caused by too strict threshold; when the data quality is high, the adaptive matching threshold is appropriately improved, which can more accurately identify potential damage and ensure the accuracy and reliability of the detection results.

[0103] S340: inputting the plurality of stress feature sets into the structure damage feature library respectively for multi-dimensional feature matching, and selecting a damage feature with a matching degree meeting the adaptive matching threshold, and outputting the plurality of damage feature sets, wherein the damage feature includes a damage type and a damage degree of the cement-based pole, and the damage type at least includes a crack, a spalling, a deformation and a corrosion.

[0104] Further, the step S340 of the present application further comprises:

[0105] S341: randomly selecting a first stress feature set and randomly selecting a first damage recognition branch in the structure damage feature library.

[0106] Specifically, first, randomly select any one of the plurality of stress feature sets as a first stress feature set, and randomly select a first damage recognition branch in the plurality of damage recognition branches of the structure damage feature library.

[0107] S342: input the first stress feature set into the first damage identification branch, perform multi-dimensional feature matching with the first standard stress feature set, if the matching degrees all satisfy the adaptive matching threshold, output the first standard damage type and the first damage degree interval as the first damage feature.

[0108] Further, the step S342 of the application further comprises:

[0109] S3421: obtain the first stress feature set, wherein the first stress feature set comprises a first abnormal stress component, a first abnormal stress proportion, a first stress change curve and a first waveform curve;

[0110] S3422: perform feature matching on the first abnormal stress component and the first standard abnormal stress component to determine a first matching degree;

[0111] S3423: if the first matching degree is greater than the adaptive matching threshold, perform feature matching on the first abnormal stress proportion and the first standard abnormal stress proportion to determine a second matching degree;

[0112] S3424: if the second matching degree is greater than the adaptive matching threshold, perform feature matching on the first stress change curve and the first standard stress change curve to determine a third matching degree;

[0113] S3425: if the third matching degree is greater than the adaptive matching threshold, perform feature matching on the first waveform curve and the first standard waveform curve to determine a fourth matching degree, if the fourth matching degree is greater than the adaptive matching threshold, output the first standard damage type and the first damage degree interval as the first damage feature.

[0114] Specifically, the first stress feature set is acquired, wherein the first stress feature set includes a first abnormal stress component, a first abnormal stress proportion, a first stress change curve, and a first waveform curve; then, the first abnormal stress component and a first standard abnormal stress component are subjected to feature matching, such as feature matching by calculating the Euclidean distance, to obtain a first matching degree; if the first matching degree is greater than the adaptive matching threshold, the first abnormal stress proportion and a first standard abnormal stress proportion are subjected to feature matching to determine a second matching degree; if the second matching degree is greater than the adaptive matching threshold, the first stress change curve and a first standard stress change curve are subjected to feature matching to determine a third matching degree; if the third matching degree is greater than the adaptive matching threshold, the first waveform curve and a first standard waveform curve are subjected to feature matching to determine a fourth matching degree; if the fourth matching degree is greater than the adaptive matching threshold, a first standard damage type (such as a crack, a spalling, corrosion, etc.) and a first damage degree interval are set as the first damage feature. If the first matching degree is less than or equal to the adaptive matching threshold, or the second matching degree is less than or equal to the adaptive matching threshold, or the third matching degree is less than or equal to the adaptive matching threshold, or the fourth matching degree is less than or equal to the adaptive matching threshold, the present matching is stopped.

[0115] S343: The first stress feature set is sequentially subjected to multidimensional feature matching with standard stress feature sets of other damage identification branches in the structural damage feature library to acquire a first damage feature set, which is added to the plurality of damage feature sets.

[0116] Specifically, the first stress feature set is sequentially subjected to multidimensional feature matching with standard stress feature sets of other damage identification branches in the structural damage feature library by using the same method, a first damage feature set satisfying the adaptive matching threshold is output, and a plurality of damage feature sets of a plurality of monitoring points are sequentially acquired. Finally, the plurality of damage feature sets are subjected to cross-set analysis to output a plurality of key damage features. Through cross-set analysis, damage features commonly supported between the plurality of monitoring points and different feature sets can be found out, signal interference and redundant features can be removed, the most representative damage features can be extracted, and the accuracy, reliability, and robustness of the cement-based electric pole defect detection are significantly improved.

[0117] S400: Concrete carbonation features and steel bar corrosion features of the target electric pole in a preset period are acquired, and the plurality of key damage features are subjected to multimodal fusion analysis according to the concrete carbonation features and the steel bar corrosion features to output an optimally matched key damage feature as a cement-based electric pole defect detection result of the target electric pole.

[0118] Further, the step S400 of the present application further includes:

[0119] S410: retrieve a sample concrete carbonation feature set, a sample steel bar corrosion feature set, and a plurality of sample damage feature sets as attribute constraints with the utility pole attribute feature and the cement-based material attribute feature, a feature constraint with the damage type and the damage degree, and a guide of the cement-based utility pole damage detection, and statistically determine a sample damage feature proportion under different sample concrete carbonation features and different sample steel bar corrosion features, set as a matching probability of a plurality of damage features, construct a sample damage feature probability distribution, and obtain a sample damage feature probability distribution set;

[0120] S420: construct a multi-modal feature fusion model, specifically including:

[0121] S4201: configure a modal feature extraction unit to perform deep feature extraction on the input stress feature, concrete carbonation feature, and steel bar corrosion feature respectively, and obtain high-dimensional feature representations of each modal;

[0122] S4202: use a three-dimensional convolutional neural network to process the concrete carbonation depth distribution feature, and extract the carbonation spatial distribution pattern;

[0123] S4203: use a time series neural network to analyze the steel bar corrosion potential change feature, and extract the corrosion development dynamic feature;

[0124] S4204: integrate the topological relationship of the stress feature, carbonation feature, and corrosion feature through a graph neural network, and establish a multi-modal feature correlation graph.

[0125] S430: calculate the weight coefficients of each modal feature using an attention mechanism, including:

[0126] calculate the attention score by the learnable query vector and the key vector of each modal feature;

[0127] perform Softmax normalization processing on the attention score to obtain the weight coefficient of each modal;

[0128] S440: weight and fuse each modal feature based on the weight coefficient to generate a fused multi-modal joint feature vector;

[0129] S450: input the sample concrete carbonation feature set, sample steel bar corrosion feature set, and sample damage feature probability distribution set into the multi-modal feature fusion model for training, specifically including:

[0130] S4501: input the fused multi-modal joint feature vector, and use the sample damage feature probability distribution as a supervision signal;

[0131] S4502: use a cross-entropy loss function to optimize the model parameters until the model converges;

[0132] S4503: Evaluate the performance of the model by the validation set, and adjust the hyperparameters of the attention mechanism;

[0133] S460: Input the stress feature, concrete carbonation feature and steel bar corrosion feature of the target power pole into the trained multi-modal feature fusion model, and perform damage feature matching analysis:

[0134] S4601: Calculate the matching score of each key damage feature in the fusion feature space;

[0135] S4602: Probability processing is performed on the matching score to generate a damage feature probability distribution;

[0136] S4603: Select the key damage feature with the highest matching score in the probability distribution as the optimal matching result;

[0137] S470: Set the optimal matching result as the cement-based power pole defect detection result of the target power pole, and output the corresponding damage type and damage degree evaluation report.

[0138] Further, the training process of the multi-modal feature fusion model further includes:

[0139] Adopting an adversarial training strategy to enhance the robustness of the model and improving the resistance of the model to signal interference by generating adversarial samples;

[0140] Introducing a transfer learning mechanism to initialize parameters using a pre-trained concrete structure damage identification model;

[0141] Setting a dynamic weight decay strategy to automatically adjust the attention weight range of each modality during training.

[0142] Specifically, in cement-based power pole defect detection, stress features, concrete carbonation features and steel bar corrosion features constitute a multi-modal damage representation system. First, based on the design parameters of the power pole (such as height, diameter, etc.) and the properties of the cement-based material (such as strength grade, mix ratio, etc.), device constraints are constructed, combined with typical damage types (cracks, spalling, exposed steel bars, etc.) and damage degree classification standards, and matching sample feature sets are retrieved from the historical case library. Through big data analysis, the frequency of occurrence of various damages under different carbonation depths and corrosion rates is calculated, and a probabilistic prior knowledge base is established.

[0143] The multi-modal feature fusion model adopts a three-level processing architecture: (1) the feature encoding layer uses a one-dimensional convolutional network to extract local features of each modality; (2) the attention fusion layer dynamically calculates the contribution degree of each modality through a multi-head attention mechanism, wherein the stress feature weight reflects the influence of mechanical load, the carbonation feature weight represents the material degradation degree, and the corrosion feature weight indicates the internal steel bar state; (3) the decision layer adopts a fully connected network to calculate the matching probability of each candidate damage. In the training process, a curriculum learning strategy is adopted, that is, first train a single modality basic model, and then gradually introduce other modalities for joint training.

[0144] The matching degree score calculation adopts Mahalanobis distance measurement, considering the covariance relationship between features:

[0145]

[0146] wherein d i represents the i-th type of damage feature, x is the input feature vector, μ i and ∑ i () are the mean and covariance matrix of the damage in the training set, respectively, T represents transposition, and finally the score is converted into a probability distribution through a Softmax function, and the damage feature corresponding to the maximum probability is selected as the detection result. Through this scheme, the damage type and degree of the cement-based power pole can be identified in real time and efficiently, and the accuracy and reliability of the defect detection of the cement-based power pole are significantly improved.

[0147] Further specifically, the method realizes accurate detection of defects of the cement-based power pole through multi-modal sensor data fusion. First, the phenolphthalein reagent method is used to measure the carbonation depth of concrete to obtain the carbonation gradient distribution in the longitudinal and radial directions of the pole body; the half-cell potential method is used to detect the steel bar corrosion state to record the potential space-time variation curve. The feature extraction network based on deep learning can automatically learn the correlation pattern of carbonation and corrosion, and the graph neural network effectively captures the complex interaction between multi-modal features. The dynamic weight distribution mechanism ensures that the optimal detection performance is maintained under different working conditions, and the finally output structured report provides a reliable basis for the health state evaluation of the power pole. Compared with single modality detection technology, the defect recognition accuracy of the method is improved by 15-20%, and the method is especially suitable for rapid and accurate detection of hidden defects of aging power poles.

[0148] It should be noted that in the cement-based utility pole defect detection, in addition to the stress data, the concrete carbonation characteristics and the steel bar corrosion characteristics are also very important auxiliary information, by combining these characteristics, the accuracy and reliability of the defect detection can be further improved. First, the concrete carbonation characteristics and the steel bar corrosion characteristics of the target utility pole within a preset period are collected, such as collecting the concrete carbonation depth data of the target utility pole within a preset period (for example, monthly or quarterly), generating a concrete carbonation feature set, the concrete carbonation depth may be related to the structural damage (such as protective layer spalling, steel bar corrosion, etc.) of the utility pole, therefore the carbonation characteristics are helpful for detecting the existence of damage; the steel bar corrosion sensor is arranged at the key position of the utility pole, and the steel bar corrosion rate data is collected. The corrosion of the steel bar of the utility pole is often accompanied by the decline of the structural strength, which is reflected in the corrosion signal; the corrosion data of the target utility pole is obtained through the steel bar corrosion sensor, and a steel bar corrosion feature set is generated, and the steel bar corrosion feature can be extracted by analyzing the corrosion rate, corrosion distribution and other parameters.

[0149] It should be noted that the stress characteristics, concrete carbonation characteristics and steel bar corrosion characteristics of the target utility pole are input into the trained multi-modal feature fusion model for damage feature matching analysis: the matching degree score of each key damage feature in the fusion feature space is calculated; the matching degree score is probabilistic processed to generate a damage feature probability distribution; the key damage feature with the highest matching degree score in the probability distribution is selected as the optimal matching result; through multi-point stress monitoring feature matching, multiple related potential damage features can be quickly found, thereby significantly improving the efficiency, accuracy and reliability of the cement-based utility pole defect detection.

[0150] Further, it also includes:

[0151] Generating a damage warning level based on the cement-based utility pole defect detection result;

[0152] Matching the damage warning level with the corresponding maintenance scheme in the maintenance strategy library;

[0153] After associating the maintenance scheme with the utility pole location information, it is pushed to the maintenance terminal.

[0154] It should be noted that a five-level warning level system is constructed, including normal, attention, warning, danger and emergency; a multi-dimensional warning index is designed: considering the influence coefficient of damage type on structure safety; introducing the ratio parameter of damage degree and critical threshold; comprehensive environmental corrosion rate factor; dynamic warning level calculation: based on damage feature calculation basic warning level; adjust the warning level combined with real-time environmental monitoring data; consider the trend change of historical detection results to correct the warning level;

[0155] Maintenance strategy intelligent matching: a strategy library containing 12 typical maintenance schemes is established; the optimal maintenance scheme is matched based on the early warning level and damage characteristics; maintenance information push: structured maintenance work order is generated; the location of the pole is located through the GIS system; and the mobile terminal of the nearest maintenance team is pushed.

[0156] Further, generating a damage early warning level based on the cement-based pole defect detection result, comprising:

[0157] Establishing a mapping relationship table of damage type, damage degree and early warning level;

[0158] According to the damage type and damage degree in the cement-based pole defect detection result, the mapping relationship table is queried;

[0159] Combined with the environmental factors of the pole, the early warning level is corrected, and the final damage early warning level is output.

[0160] It should be noted that the damage-early warning mapping knowledge base is constructed: the corresponding relationship between 8 typical damage modes and early warning levels is defined; the damage degree grading standard is established; the environmental factor correction module: environmental temperature and humidity, salt fog concentration data are collected; the environmental acceleration factor is calculated; the environmental correction coefficient matrix is designed; the early warning level decision model: fuzzy reasoning system is used to process uncertain information; the grade adjustment mechanism based on case reasoning is constructed; early warning verification and optimization: setting up expert review link; establishing early warning effect feedback closed loop; regularly updating early warning rule library; early warning information visualization: generating three-dimensional early warning situation map; providing multi-dimensional early warning analysis report.

[0161] It should be understood that the above embodiments are one or more embodiments of the present application, and there are many other embodiments and variations of the present application based on the present application; the deformation and modification of the present application by ordinary skilled in the art without making pioneering innovation, all belong to the protection scope of the present application.

Claims

1. A cement-based pole defect detection method based on big data analysis, characterized in that, The method comprises: Based on the predetermined monitoring point distribution of distributed sensing network, through the distributed sensing network to the target electric pole for continuous monitoring, obtain a plurality of stress data sets within a predetermined period; Stress feature extraction is performed on the plurality of stress data sets to obtain a plurality of stress feature sets; The plurality of stress feature sets are respectively input into the structure damage feature library for multi-dimensional feature matching, a plurality of damage feature sets satisfying the adaptive matching threshold are selected, and cross set analysis is performed on the plurality of damage feature sets, and a plurality of key damage features are output; Collect the concrete carbonation characteristics and steel corrosion characteristics of the target electric pole within a predetermined period, and perform multi-modal fusion analysis on the plurality of key damage features according to the concrete carbonation characteristics and steel corrosion characteristics, and output the optimal matching key damage feature as the cement-based electric pole defect detection result of the target electric pole.

2. A cement-based pole defect detection method based on big data analysis according to claim 1, characterized in that, Based on the predetermined monitoring point distribution of distributed sensing network, through the distributed sensing network to the target electric pole for continuous monitoring, obtain a plurality of stress data sets, comprising: Obtain the predetermined monitoring point of the target electric pole, wherein the predetermined monitoring point at least includes the middle part of the pole body, the bottom of the pole body and the pole body connection; Distributed sensing network is constructed by arranging optical fiber stress sensors at the middle part of the pole body, the bottom of the pole body and the pole body connection; According to the predetermined monitoring frequency, the target electric pole is continuously monitored through the distributed sensing network, and a plurality of stress data sets of a plurality of monitoring points within a predetermined period are obtained.

3. A cement-based pole defect detection method based on big data analysis according to claim 2, characterized in that, Stress feature extraction is performed on the plurality of stress data sets to obtain a plurality of stress feature sets, comprising: Signal interference analysis and filtering denoising processing are performed on the plurality of stress data sets to obtain a plurality of signal interference ratios and a plurality of standard stress data sets; Configure stress feature index, wherein the stress feature index at least includes abnormal stress component, abnormal stress proportion, stress change curve and waveform curve; According to the stress feature index, stress feature extraction is performed on the plurality of standard stress data sets to obtain a plurality of stress feature sets of the plurality of monitoring points.

4. A cement-based pole defect detection method based on big data analysis according to claim 3, characterized in that, The plurality of stress feature sets are respectively input into the structure damage feature library for multi-dimensional feature matching, a plurality of damage feature sets satisfying the adaptive matching threshold are selected, comprising: Based on the electric pole attribute characteristics, the cement-based material attribute characteristics and the stress feature index, a structure damage feature library is constructed; The mean value of the plurality of signal interferences is calculated to determine the comprehensive signal-to-noise ratio, the ratio of the comprehensive signal-to-noise ratio to the standard signal-to-noise ratio is calculated, and the threshold compensation coefficient is obtained; According to the threshold compensation coefficient, the initial similarity comparison threshold is compensated to obtain the adaptive matching threshold; The plurality of stress feature sets are respectively input into the structure damage feature library for multi-dimensional feature matching, and the damage features satisfying the adaptive matching threshold are selected, and the plurality of damage feature sets are output, wherein the damage features include the damage type and damage degree of the cement-based electric pole, and the damage type at least includes crack, peeling, deformation and corrosion.

5. A cement-based pole defect detection method based on big data analysis according to claim 4, characterized in that, Based on the electric pole attribute characteristics, the cement-based material attribute characteristics and the stress feature index, a structure damage feature library is constructed, comprising: Using the attribute characteristics of the utility pole and the attribute characteristics of the cement-based material as attribute constraints, the stress characteristic index as condition constraints, and the cement-based utility pole damage detection based on stress analysis as a guide, information retrieval is performed based on big data to obtain multiple sample stress feature sets, and the damage type and damage degree of cement-based utility poles under different sample stress feature sets are labeled to obtain multiple sample damage types and multiple sample damage degrees. Clustering is performed on the multiple sample damage types and multiple sample damage degrees based on the same damage type and damage degree interval to determine multiple standard damage types, wherein each standard damage type is identified with multiple damage degree intervals; A first standard damage type and a first damage degree range of the first standard damage type are randomly selected. The stress feature sets of the multiple samples are screened according to the first standard damage type and the first damage degree range to determine multiple first sample stress feature sets. High-frequency feature analysis is performed on the plurality of first sample stress feature sets to determine a first standard stress feature set, wherein the first standard stress feature set includes a first standard abnormal stress component, a first standard abnormal stress proportion, a first standard stress change curve, and a first standard waveform curve; Establish a first mapping association between the first standard stress feature set and the first standard damage type and the first damage degree range, and construct a first damage identification branch based on the first mapping association; Multiple damage identification branches are obtained by sequential analysis, and the structural damage feature library is constructed based on the multiple damage identification branches.

6. A cement-based pole defect detection method based on big data analysis according to claim 5, characterized in that, The multiple stress feature sets are respectively input into the structural damage feature library for multi-dimensional feature matching, and damage features that meet the adaptive matching threshold are selected. The multiple damage feature sets are then output, including: A first stress feature set is randomly selected, and a first damage identification branch is randomly selected from the structural damage feature library; The first stress feature set is input into the first damage identification branch and multi-dimensional feature matching is performed with the first standard stress feature set. If the matching degree meets the adaptive matching threshold, the first standard damage type and the first damage degree range are output as the first damage feature. The first stress feature set is sequentially matched with the standard stress feature sets of other damage identification branches in the structural damage feature library in a multi-dimensional feature matching process to obtain the first damage feature set, which is then added to the plurality of damage feature sets.

7. A cement-based pole defect detection method based on big data analysis according to claim 6, characterized in that, The first stress feature set is input into the first damage identification branch, and multi-dimensional feature matching is performed with the first standard stress feature set. If the matching degree satisfies the adaptive matching threshold, the first standard damage type and the first damage degree range are output as the first damage feature, including: Obtain the first stress feature set, wherein the first stress feature set includes a first abnormal stress component, a first abnormal stress proportion, a first stress change curve, and a first waveform curve; Perform feature matching on the first abnormal stress component and the first standard abnormal stress component to determine the first matching degree; If the first matching degree is greater than the adaptive matching threshold, the first abnormal stress ratio and the first standard abnormal stress ratio are matched in feature, and a second matching degree is determined; If the second matching degree is greater than the adaptive matching threshold, the first stress change curve and the first standard stress change curve are matched in feature, and a third matching degree is determined; If the third matching degree is greater than the adaptive matching threshold, the first waveform curve and the first standard waveform curve are matched in feature, and a fourth matching degree is determined, and if the fourth matching degree is greater than the adaptive matching threshold, the first standard damage type and the first damage degree interval are set as the first damage feature.

8. A cement-based pole defect detection method based on big data analysis according to claim 1, characterized by, According to the concrete carbonation characteristics and the reinforcement corrosion characteristics, the plurality of key damage features are analyzed in multi-modal fusion, and the optimal matching key damage feature is set as the cement-based electric pole defect detection result of the target electric pole, including: The sample concrete carbonation feature set, the sample reinforcement corrosion feature set and the plurality of sample damage feature sets are retrieved and obtained by taking the electric pole attribute feature and the cement-based material attribute feature as attribute constraints, taking the damage type and the damage degree as feature constraints, and taking the cement-based electric pole damage detection as guidance; A multi-modal feature fusion model is constructed, and the stress feature, the concrete carbonation feature and the reinforcement corrosion feature are taken as input modalities; The attention mechanism is used to calculate the weight coefficient of each modal feature; Based on the weight coefficient, the modal features are weighted and fused to calculate the matching degree score of each key damage feature; The key damage feature with the highest matching degree score is selected as the optimal matching result, and is set as the cement-based electric pole defect detection result of the target electric pole.

9. A cement-based pole defect detection method based on big data analysis according to claim 8, characterized in that, Further comprising: Based on the cement-based electric pole defect detection result, a damage early warning level is generated; According to the damage early warning level, a corresponding maintenance scheme in the maintenance strategy library is matched; The maintenance scheme is associated with the electric pole position information and pushed to the maintenance terminal.

10. A cement-based pole defect detection method based on big data analysis according to claim 9, characterized in that, Based on the cement-based electric pole defect detection result, a damage early warning level is generated, including: A mapping relationship table of damage type, damage degree and early warning level is established; According to the damage type and the damage degree in the cement-based electric pole defect detection result, the mapping relationship table is queried; Combined with the environmental factors of the electric pole, the early warning level is corrected, and the final damage early warning level is output.

Images