A method and system for evaluating the weather resistance and predicting the service life of a super-amphiphobic insulator

Abstract

This invention provides a method and system for weather resistance assessment and lifetime prediction of super-hydrophobic insulators, comprising: applying temperature and humidity, ultraviolet irradiation, contamination deposition, and multi-physics coupling of electric fields to accelerate the aging of super-hydrophobic insulators; simultaneously collecting multi-source time-series degradation data during the aging process, the data covering surface hydrophobicity, micro / nano morphology, low surface energy component mobility, and coating interface bonding strength parameters; based on the multi-source time-series degradation data, extracting a degradation state vector that integrates multi-scale degradation features, the features including at least an integrity index reflecting the relative change in micro / nano structure coverage and roughness, a migration index based on the time-varying ratio of low surface energy component concentration, and a time-series damage factor weighted by the rate of change of multiple degradation parameters; inputting the degradation state vector into a pre-trained time-series deep learning model to output the remaining lifetime and failure risk level of the super-hydrophobic insulator.

Description

Technical Field

[0001] This invention belongs to the technical field of weather resistance assessment and life prediction of power equipment materials, specifically relating to a method and system for weather resistance assessment and life prediction of super-dual-diffusivity insulators. Background Technology

[0002] Ultra-high voltage (UHV) power transmission systems, as the core support for the "West-to-East Power Transmission," "Smart Grid," and Global Energy Internet strategies, have become critical infrastructure for ensuring national energy security, optimizing energy resource allocation, and supporting high-quality economic and social development. Their coverage spans diverse and complex geographical areas, including high-altitude regions, coastal areas, industrial zones, and mountainous areas. The operating environment is characterized by drastic temperature and humidity fluctuations, intense ultraviolet radiation, diverse types of pollution (industrial dust, sea salt particles, vegetation secretions, etc.), and continuous electric field stress. The external insulation components of transmission and transformation equipment (especially insulators) are exposed to such complex environments for extended periods. Their performance stability directly determines the safety and reliability of the power grid operation, and the coupling effect of pollution deposition and humid environments has become the primary cause of external insulation failure.

[0003] When contaminants deposited on the surface of insulators absorb moisture and dissolve under humid conditions such as rain or condensation, they form a conductive film, significantly reducing the insulation resistance of the insulator surface. This leads to increased leakage current, which in turn triggers partial discharge and eventually develops into a flashover accident. Flashover is characterized by its sudden onset, wide impact, and long recovery time. It not only causes large-scale power outages but also results in cascading losses such as equipment damage and soaring maintenance costs. In severe cases, it can even trigger a chain reaction of power grid failures, causing incalculable impacts on industrial production, residential life, and public safety. Therefore, improving the flashover resistance of insulators and accurately assessing their weather resistance and remaining lifespan in complex environments have become core technical challenges that urgently need to be addressed in the field of power grid operation and maintenance.

[0004] To address the risk of flashover, various protective and maintenance measures have been developed in existing technologies, but all of them have significant limitations. Regular manual or mechanical cleaning is the traditional mainstream method. However, it not only consumes a lot of manpower and material resources, but is also limited by factors such as terrain and climate, making it difficult to achieve timely cleaning of the entire line. Insulators still face the risk of flashover within the cleaning interval. Adjusting the creepage distance of insulators can improve the anti-pollution capability to a certain extent, but it will lead to a significant increase in the size and weight of the insulators, which not only increases the cost of equipment manufacturing, transportation and installation, but may also be limited by the space of the line corridor, making it impossible to flexibly apply to the renovation of existing lines. Although the application of composite insulators improves mechanical properties and aging resistance, long-term exposure to ultraviolet radiation, ozone oxidation and pollution corrosion environment will easily cause cracking and powdering of their surface, resulting in a gradual loss of hydrophobicity. Coating with ordinary hydrophobic coatings is a common improvement solution in recent years, but its contact angle is usually only 100°–110°, its self-cleaning ability is limited, and the low surface energy components in the coating are prone to migration, volatilization or wear during long-term use, resulting in rapid decay of hydrophobicity. Frequent recoating and maintenance are required, which not only increases the workload of operation and maintenance, but may also cause safety hazards due to untimely maintenance. More importantly, existing technologies lack systematic evaluation methods for the stability of micro-nano structures, surface function degradation mechanisms, and life evolution laws of insulators under multi-stress coupling environments of temperature, humidity, ultraviolet radiation, pollution, and electric fields. This makes it impossible to provide accurate life prediction and maintenance decision support for equipment operation and maintenance, resulting in the external insulation design still needing to be based on conservative hydrophilic surface redundancy configuration, which further exacerbates the problems of bulky equipment and frequent maintenance. Summary of the Invention

[0005] To address the shortcomings and deficiencies of existing technologies, this invention provides a method and system for weather resistance assessment and lifetime prediction of super-double hydrophobic insulators. The aim is to solve the problem that existing technologies lack systematic, in-situ dynamic assessment and accurate prediction of the microstructural stability, surface functional degradation mechanism, and lifetime evolution law of super-double hydrophobic materials under complex multi-stress coupling environments.

[0006] The core of this invention lies in constructing a complete technical system integrating multi-physics field coupled accelerated aging, in-situ multi-modal synchronous monitoring, multi-scale feature fusion, intelligent time-series prediction, and closed-loop decision-making. Specifically, firstly, an environmental simulation chamber integrating temperature and humidity control, ultraviolet irradiation, contamination settling, and high-voltage electric field modules is used to apply multi-stress synergistic accelerated aging of the super-hydrophobic insulator under simulated real service conditions. During this process, an in-situ characterization platform integrating high-speed imaging, laser confocal microscopy, surface chemical analysis modules, and micro-force sensors is used to simultaneously collect multi-source time-series degradation data, including surface hydrophobic properties (such as contact angle and roll-off angle), self-cleaning efficiency, micro / nano morphology, changes in low surface energy component concentration, and coating interface bonding strength, without interrupting the aging process.

[0007] To address the challenge of fusing multi-source heterogeneous data, this invention performs fusion preprocessing on the collected data, including outlier removal, time alignment, dynamic baseline correction, and cross-modal normalization. Based on the preprocessed data, an innovative fusion feature vector is extracted and constructed to comprehensively reflect the multi-scale degradation state of the material. This vector not only includes the degradation rate of macroscopic properties, but more importantly, it introduces an "integrity index" reflecting the synergistic change of micro / nanostructure coverage and roughness, a "migration index" characterizing the degree of loss of low surface energy components over time, and a "time-series damage factor" that integrates the weighted accumulation of the change rates of multiple parameters. These features deeply characterize the unique degradation mode of the superhydrophobic coating from the perspectives of microstructure, chemical composition, and time-series cumulative effects.

[0008] Subsequently, the constructed degradation state vector is input into a pre-trained long short-term memory network model incorporating an attention mechanism. This model can capture long-term dependencies and key time nodes in the degradation process, thereby outputting the remaining service life and failure risk level of the super-dual-diffusivity insulator. To improve engineering practicality, this invention also designs a closed-loop decision-making logic based on the prediction results: determining the health status, issuing warnings, or generating maintenance strategies based on the remaining service life and risk level; when the prediction uncertainty is high, supplementary experiments can be automatically triggered to update the model, achieving self-evolution of prediction capabilities.

[0009] Accordingly, the present invention also provides a system for performing the above method, the core of which includes an environmental simulation chamber for realizing multi-stress coupled loading, a characterization platform for realizing in-situ synchronous monitoring, and a data processing unit for data fusion and intelligent prediction.

[0010] Through the above-mentioned technical solution, this invention realizes a full-chain, quantitative evaluation of the weather resistance performance of super-double hydrophobic insulators, from "phenomenon monitoring" to "mechanism characterization" and then to "life prediction," providing key technical support for the reliable application and intelligent operation and maintenance of super-double hydrophobic materials in the field of power external insulation.

[0011] The specific technical solution adopted by this invention to solve its technical problem is as follows:

[0012] A method for weather resistance assessment and life prediction of super-diffuse insulators includes:

[0013] Applying temperature and humidity, ultraviolet radiation, dirt deposition, and multi-physics coupling of electric fields to super-dual-hydrophobic insulators accelerates aging.

[0014] Multi-source time-series degradation data were collected simultaneously during the aging process. The data covered parameters such as surface hydrophobicity, micro-nano morphology, low surface energy component mobility, and coating interface bonding strength.

[0015] Based on the multi-source temporal degradation data, a degradation state vector that integrates multi-scale degradation features is extracted. The features include at least an integrity index reflecting the relative change in micro / nano structure coverage and roughness, a migration index based on the time-varying ratio of low surface energy component concentration, and a temporal damage factor that is a weighted accumulation of the change rates of multiple degradation parameters.

[0016] The degradation state vector is input into a pre-trained temporal deep learning model, which outputs the remaining service life and failure risk level of the super double-diffuse insulator.

[0017] Furthermore, the parameters for the multi-physics coupled accelerated aging are set in conjunction with the target service environment and acceleration factor. The core parameter ranges include: a temperature and humidity cycling range of -10℃ to 40℃, a relative humidity of 30% to 95%, and an ultraviolet irradiation intensity of 0.63 to 1.20 W / m at a narrow band control point of 340nm. 2 / nm, with a fouling settling rate of 0.1 to 0.5 mg / cm³. 2 • day, applied voltage is ±50kV.

[0018] Furthermore, the integrity index is the product of the micro / nano structure coverage and the attenuation coefficient of the root mean square roughness after aging relative to the initial root mean square roughness; the migration index is the ratio of the real-time migration rate of low surface energy components during aging to the initial migration rate; and the weighting coefficient of the time-series damage factor is automatically determined from the training samples through principal component analysis.

[0019] Furthermore, before extracting multi-scale degradation features, multi-source temporal degradation data are fused and preprocessed. The preprocessing includes time alignment, data smoothing, trend correction, and cross-modal normalization. The cross-modal normalization introduces a nonlinear weighting factor based on the parameter change rate for adjustment.

[0020] Furthermore, the temporal deep learning model is a long short-term memory network model that includes an attention mechanism.

[0021] Furthermore, data acquisition employs a strategy of combining the master sample with control samples from the same batch: the master sample is used for monitoring electric field and wet contamination coupling aging and electrical characteristics, while the control samples from the same batch are used for testing contact angle, self-cleaning efficiency, and interfacial bonding strength.

[0022] Furthermore, after outputting the remaining service life and failure risk level, the following decision-making logic is applied: if the remaining service life is not less than the first set threshold and the failure risk level is low, it is judged to be in a healthy state; if the remaining service life is not less than the second set threshold and less than the first set threshold, or the failure risk level is medium, an early warning is triggered; if the remaining service life is less than the second set threshold or the failure risk level is high, a failure mechanism diagnosis report and maintenance strategy are generated in combination with the dominant degradation mode; if the prediction uncertainty exceeds the preset threshold, a secondary accelerated aging experiment is automatically started, the local model is retrained after supplementing the data, and the prediction results are updated.

[0023] Furthermore, in the multi-source temporal degradation data:

[0024] The migration rate of the low surface energy components is obtained by periodically detecting the change rate of fluorine or silicon surface concentration using a surface chemical analysis module.

[0025] The coating interface bonding strength is obtained by applying a vertical peeling force using a micro-force sensor and recording the critical desorption load.

[0026] All data is stored using a unified timestamp.

[0027] Furthermore, when training the temporal deep learning model, the Huber loss function is used as the optimization objective, and an early stopping mechanism is introduced to prevent overfitting.

[0028] And, a weather resistance assessment and life prediction system for ultra-diffuse insulators, for performing the method described above, including:

[0029] The multiphysics environment simulation chamber is configured to simultaneously apply temperature and humidity stress, ultraviolet radiation stress, pollution settlement stress and electric field stress to the super-dual-hydrophobic insulator.

[0030] The in-situ characterization platform is configured to simultaneously acquire multi-source time-series degradation data during the aging process, including a high-speed camera system, a laser confocal microscope, a surface chemical analysis module, a micro-force sensor, and a six-axis robotic arm. The surface chemical analysis module is equipped with a sealed sampling and vacuum transfer component to achieve quasi-in-situ analysis.

[0031] The data processing unit is communicatively connected to the in-situ characterization platform and is configured to perform data fusion preprocessing, multi-scale degradation feature extraction, degradation state vector construction, and lifetime prediction model calculation.

[0032] And a computer device including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the method described above.

[0033] A non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method described above.

[0034] Compared with the prior art, the present invention and its preferred embodiments have at least the following beneficial effects:

[0035] First, this invention breaks through the limitations of traditional evaluation methods that are singular and static. By constructing an accelerated aging environment with multi-physical field coupling of temperature, humidity, ultraviolet light, pollution, and electric field, it can more realistically and comprehensively simulate the service conditions of super-dual-hydrophobic insulators under complex natural conditions, providing a practical technical basis for evaluating their long-term weather resistance.

[0036] Secondly, by integrating multiple in-situ characterization methods and achieving simultaneous monitoring, this invention can simultaneously acquire multi-dimensional and multi-scale time-series degradation data, ranging from macroscopic hydrophobic properties to microstructural morphology, and from surface chemical composition to interfacial mechanical strength, without interrupting the aging process or damaging the sample surface. This in-situ, multi-modal data acquisition method effectively avoids interference from sampling tests, ensures the integrity and authenticity of the data chain, and provides unprecedented data support for a deeper understanding of the degradation mechanism of superhydrophobic materials.

[0037] Furthermore, addressing the challenge of fusing multi-source heterogeneous data, this invention employs a preprocessing workflow encompassing outlier handling, time alignment, and cross-modal normalization. It also innovatively constructs a multi-scale degradation state vector that integrates macroscopic performance degradation rate, microstructural integrity, chemical component mobility, and temporal cumulative damage factors. This feature engineering method transforms the complex physicochemical degradation process into machine-learnable quantitative features, profoundly revealing the unique functional degradation patterns of superhydrophobic coatings.

[0038] Furthermore, by employing a temporal deep learning model incorporating an attention mechanism to learn and predict the aforementioned feature vectors, this invention can effectively capture long-term dependencies and key inflection points in the degradation process, thereby achieving accurate and intelligent prediction of remaining service life and failure risk level, changing the previous situation of relying on empirical formulas or acceleration rates for rough estimation.

[0039] Finally, the complete technical chain formed by this invention, from "multi-stress accelerated aging" to "in-situ synchronous characterization", then to "multi-feature fusion prediction" and "closed-loop decision update", not only realizes the systematic evaluation of the weather resistance of super-dual-hydrophobic insulators, but also elevates it to a predictable and manageable intelligent level, providing reliable technical tools and decision-making basis for the precise operation and maintenance of power equipment, life cycle cost optimization, and the research and development and application of new super-dual-hydrophobic materials. Attached Figure Description

[0040] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments:

[0041] Figure 1 This is a schematic diagram of the overall structure of the super-dual-diffusivity insulator weather resistance assessment and life prediction system according to an embodiment of the present invention;

[0042] Figure 2 This is a schematic diagram of the internal structure of the multimodal environment simulation chamber according to an embodiment of the present invention;

[0043] Figure 3 This is a schematic diagram showing the changes of key functional parameters of the present invention with accelerated aging time; in the figure, (a) is a graph showing the degradation trend of contact angle and roll-off angle, and (b) is a graph showing the changes in self-cleaning efficiency and interface strength.

[0044] Figure 4 This is a schematic diagram of data analysis of multimodal degradation features in an embodiment of the present invention; in the figure, (a) is a correlation matrix diagram of degradation features, and (b) is a PCA dimensionality reduction projection diagram (color represents aging time).

[0045] Figure 5 This is a schematic diagram comparing the lifetime prediction results of the improved LSTM-Attention model in an embodiment of the present invention; in the figure, (a) is a comparison diagram of the predicted value and the actual value, and (b) is a distribution diagram of the prediction error.

[0046] Figure 6 This is a schematic diagram of the failure risk level assessment results of the super-double condenser insulator according to an embodiment of the present invention; in the figure, (a) is the evolution diagram of failure risk probability over time, and (b) is the decision logic diagram of risk level. Detailed Implementation

[0047] To make the features and advantages of the present invention more apparent and understandable, specific embodiments are described below in detail:

[0048] It should be noted that the following detailed descriptions are exemplary and intended to provide further explanation of this application. Unless otherwise specified, all technical and scientific terms used in this specification have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains.

[0049] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the exemplary embodiments according to this application. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0050] To address the shortcomings of existing technologies, this invention provides a weather resistance assessment and lifetime prediction system for superhydrophobic insulators. Its purpose is to solve the problem of the lack of a systematic assessment method for the stability of micro / nanostructures, surface functional degradation mechanisms, and lifetime evolution under multi-stress coupling in superhydrophobic materials. It includes a multimodal environment simulation chamber and an in-situ characterization platform to construct accelerated aging environments under multi-stress coupling of temperature and humidity, ultraviolet radiation, pollution, and electric fields. It simultaneously collects six types of data: contact angle, roll-off angle, self-cleaning efficiency, micro / nano morphology, low surface energy component mobility, and interfacial bonding strength. After fusion and preprocessing, an 8-dimensional degradation state vector is constructed and input into an improved LSTM-Attention model to predict the remaining lifetime and failure risk level. This application enables a systematic assessment and accurate lifetime prediction of the micro / nanostructure stability and functional degradation mechanisms of superhydrophobic insulators under complex service environments.

[0051] The implementation of the solution includes the following steps:

[0052] Step S1: Deploy the multimodal environment simulation chamber (1) and the in-situ characterization platform (2) around the super-dual hydrophobic insulator (3) to be tested, and construct a multi-physics field coupling accelerated aging environment with adjustable temperature and humidity, ultraviolet radiation, dirt deposition and electric field stress;

[0053] Step S2: In a multi-physics coupling environment, configure environmental parameters and online sensing modules, and synchronously collect six types of raw data of the super-double hydrophobic insulator (3) during the aging process: contact angle, roll-off angle, self-cleaning efficiency, surface micro-nano morphology, low surface energy component migration rate and interface bonding strength.

[0054] Step S3: Perform fusion preprocessing on the raw data, including outlier removal, time alignment, sliding window smoothing, dynamic baseline correction, and cross-modal normalization;

[0055] Step S4: Construct a multi-scale degradation feature set based on the preprocessed data to form an 8-dimensional degradation state vector;

[0056] Step S5: Input the 8-dimensional degradation state vector into the improved LSTM-Attention lifetime prediction model and output the remaining lifetime and failure risk level of the super double sparse insulator (3).

[0057] The multimodal environment simulation chamber (1) and in-situ characterization platform (2) include a temperature and humidity control chamber (4), an adjustable ultraviolet array light source (5), an aerosol contaminant generator (6), a high-voltage DC electrode (7), a high-speed camera system (8), a laser confocal microscope (9), a surface chemical analysis module (10) (used for surface component testing such as XPS or ATR-FTIR, and equipped with a sealed sampling / vacuum transfer assembly to achieve quasi-in-situ analysis), a micro-force sensor (11), a six-axis robotic arm (12), and an edge computing unit (13); temperature The wet control chamber (4) is connected to the end of the six-axis robotic arm (12) via a guide rail. The ultraviolet array light source (5) is embedded in the top of the inner wall of the chamber. The aerosol contaminant generator (6) is connected to the side wall of the chamber via a pipe. The high-voltage DC electrodes (7) are symmetrically arranged at both ends of the insulator. The high-speed camera system (8) and the laser confocal microscope (9) are integrated on the liftable bracket. The surface chemical analysis module (10) and the micro-force sensor (11) are installed on the micro positioning platform. All sensing and execution units communicate with the edge computing unit (13) via the industrial bus (14).

[0058] The specific process of step S1 is as follows: Start the six-axis robotic arm (12) and put the super-double-hydrophobic insulator (3) to be tested into the center position of the temperature and humidity control chamber (4); after closing the chamber door, set the temperature and humidity cycle curve to 40℃ / 95%RH←→-10℃ / 30%RH and the ultraviolet irradiance intensity (measured at the narrow band control point at 340nm) to 0.63-1.20 W / m 2 / nm@340nm, with a fouling settling rate of 0.1-0.5 mg / cm³. 2 ·day and applied voltage are The multi-physics field collaborative loading program is initiated to allow the super-double-hydrophobic insulator (3) to undergo an accelerated aging process under simulated service conditions. In this embodiment, the pose control accuracy of the six-axis robotic arm (12) is ±0.05mm. After positioning, it is calibrated by a laser rangefinder to ensure that the center of the super-double-hydrophobic insulator (3) under test is aligned with the center of the temperature and humidity control cavity (4) and the test area of ​​the multi-mode monitoring equipment.

[0059] The online sensing module configured in step S2 includes: a high-speed camera system (8) is used to measure the contact angle / roll-off angle in conjunction with an automatic droplet dispenser, and the static contact angle is collected once every 24 hours. Contact angle with forward / reverse Roll angle Self-cleaning efficiency was determined by the tilted platform method. Calculations were performed using a standard dust spraying followed by water rinsing experiment. The standard dust used is kaolin (particle size 1-5μm) as specified in IEC 60507, and the spraying standard is 0.1g / cm³. 2The density of the material uniformly covers the test surface of the insulator. The water rinsing conditions are: water pressure 0.1 MPa, rinsing time 30 s, and rinsing direction at a 45° angle to the insulator surface. The surface micro-nano morphology is scanned every 72 hours using a laser confocal microscope (9), and the roughness is extracted. With structural coverage Low surface energy component mobility The surface concentration change rate of fluorine or silicon was periodically detected by an X-ray photoelectron spectroscopy probe (10); the interfacial bonding strength was also measured. A vertical peeling force was applied by a micro-force sensor (11) and the critical desorption load was recorded. The test area was selected as the central area of ​​the insulator surface (within a diameter of 10 mm), the peeling speed was set to 10 μm / s, and the loading method was to increase the load at a uniform rate until the coating separated from the substrate. All raw data were temporarily stored in the buffer area of ​​the edge computing unit (13) with a unified timestamp.

[0060] The specific process of step S3 is as follows:

[0061] Outlier removal: using an improved method The criteria, combined with sliding window mean filtering, remove data points that deviate from the local mean by more than three times the dynamic standard deviation.

[0062] Time alignment: Based on the aging start time, the six types of asynchronous sampled data are linearly interpolated and aligned to a unified time grid;

[0063] Sliding window smoothing: Apply a Savitzky-Golay filter with a window length of 5 and a second-order polynomial to smooth each type of data;

[0064] Dynamic baseline correction: An exponential decay model is used for contact angle and roll-off angle parameters. To fit the long-term trend, subtract the trend term from the original sequence;

[0065] Cross-modal normalization: Min-Max normalization was performed on six different types of data to the [0,1] interval, and weighting factors were introduced. Nonlinear compression is performed, where For the first The current rate of change of the class parameter, and As an empirical constant, the method for determining it for reference is as follows: based on aging test data of 10 different super-diffuse insulators, the parameter value that minimizes the cross-modal data fusion error is obtained by optimizing it through the grid search method.

[0066] The specific process of step S4 is as follows:

[0067] Macroscopic performance degradation characteristics: Extracting the contact angle decay rate Roll-off angle growth rate Self-cleaning efficiency decline rate ;

[0068] Microstructure degradation characteristics: Calculation of micro / nanostructure integrity index from laser confocal images ;

[0069] Chemical composition degradation characteristics: defining a low surface energy migration index ;

[0070] Mechanical interface degradation characteristics: Extracting the slope of interface bond strength attenuation. Time-series cumulative damage characteristics: Constructing cumulative aging factors The weight Determined by principal component analysis;

[0071] Feature standardization and dimensionality reduction: , , , , , , , Eight features were standardized using Z-score, and the eight-dimensional vectors with the highest information gain were selected as the degenerate state vectors by mutual information entropy screening. .

[0072] The improved LSTM-Attention lifetime prediction model above adopts an architecture of "8 input layers - two-layer LSTM (64 units per layer) - Attention weight layer - single output layer". The calculation formula for the Attention mechanism is as follows: , , ,in For LSTM at time The hidden state, , , For learnable parameters, The context vector is used; the output layer uses a linear activation function to predict the remaining lifetime. The model training uses the Huber loss function and introduces an early stopping mechanism to prevent overfitting. In this embodiment, the early stopping trigger condition is: when the validation set loss increases for 5 consecutive iterations, or the loss fluctuation is less than 0.1% for 3 consecutive iterations, the model training is terminated and the current optimal model parameters are saved.

[0073] After the model outputs the remaining useful life and failure risk level, the following decision is made: If the prediction... Set threshold And if the risk level is low, then it is judged as a healthy state; if If the risk level is medium, an alert will be triggered and a shorter testing cycle will be recommended; if If the risk level is high, then combine the degenerate state vector. The system generates a failure mechanism diagnosis report based on the dominant degradation mode and recommends maintenance strategies. When the prediction uncertainty exceeds 15%, a secondary accelerated aging experiment is automatically initiated. After supplementing with 5 sets of new data, the local model is retrained and the prediction results are updated.

[0074] This invention provides a specific implementation of a weather resistance assessment and life prediction system for ultra-diffuse insulators, combined with... Figures 1 to 6 The system provides a complete description of its structure, operation process, and technical details. In practical applications, this system can be deployed in high-voltage transmission line operation and maintenance laboratories or materials aging research centers to accurately assess the service performance and lifespan prediction of novel ultra-dual-hydrophobic coated insulators under complex environments.

[0075] like Figure 1 As shown, the system in this embodiment consists of a multimodal environment simulation chamber (1) and an in-situ characterization platform (2) forming the main framework. The two work together to achieve accelerated aging and real-time monitoring of the super-dual-hydrophobic insulator (3). The multimodal environment simulation chamber (1) integrates a temperature and humidity control chamber (4), an adjustable ultraviolet array light source (5), an aerosol pollution generator (6), and a high-voltage DC electrode (7). The in-situ characterization platform (2) includes a high-speed camera system (8), a laser confocal microscope (9), an X-ray photoelectron spectroscopy probe (10), a micro-force sensor (11), a six-axis robotic arm (12), and an edge computing unit (13). All sensing and execution units communicate with the edge computing unit (13) through an industrial bus (14) to form a closed-loop control and data acquisition system.

[0076] In the specific operation process, the first step S1 is to start the six-axis robotic arm (12) and automatically load the super-dual-hydrophobic insulator (3) to be tested into the center position of the temperature and humidity control chamber (4), and complete the precise positioning through the guide rail; then close the chamber door and set the temperature and humidity cycle curve to 40℃ / 95%RH. -10℃ / 30%RH, UV irradiance (measured at a narrow band control point of 340nm) is set to 0.89W / m 2 / nm@340nm, the fouling settling rate is 0.3mg / cm³. 2 ·day, applied voltage is After the multi-physics field collaborative loading program is started, the super-damp insulator (3) begins to accelerate the aging process in a multi-stress coupling environment simulating real service conditions. During this stage, the temperature and humidity control cavity (4) maintains dynamic temperature and humidity changes, the ultraviolet array light source (5) continuously irradiates the surface to induce photo-oxidation reaction, the aerosol pollution generator (6) sprays simulated pollution particles into the cavity at a set rate, and the high-voltage DC electrode (7) applies electric field stress to both ends of the insulator to induce partial discharge effect.

[0077] After entering step S2, the online sensing module synchronously collects six types of raw data. Specifically, the high-speed camera system (8), in conjunction with the automatic droplet dispenser, measures the static contact angle every 24 hours. Contact angle with forward / reverse The rolling angle was determined using the tilting platform method. Self-cleaning efficiency The formula was calculated based on a standard dust spraying followed by water rinsing experiment. ,in For the initial adhering dust quality, To measure the residual dust quality after rinsing; a laser confocal microscope (9) was used to perform a three-dimensional scan of the insulator surface every 72 hours to extract the root mean square roughness. With micro / nano structure coverage The surface chemical analysis module (10) performs quasi-in-situ surface composition testing (such as XPS or ATR-FTIR) through a closed sampling / vacuum transfer assembly to obtain the surface concentration change rate of low surface energy components such as fluorine (F) or silicon (Si), thereby characterizing the mobility of low surface energy components. The environmental control requirements for vacuum transfer are: vacuum degree ≤ 10 Pa, transfer time ≤ 30 s, to avoid oxidation of low surface energy components upon contact with air. The micro-force sensor (11) applies a vertical peeling force to the sample surface through a micro-positioning platform, recording the critical desorption load to reflect the interfacial bonding strength. All the above raw data have a unified timestamp and are temporarily stored in the cache of the edge computing unit (13).

[0078] Self-cleaning efficiency It is a dimensionless ratio, with a value ranging from 0 to 1.

[0079] Considering that self-cleaning experiments and interface bonding strength tests may cause local disturbances to the samples, this embodiment can adopt a "master sample + control sample from the same batch" strategy: the master sample is used for electric field / wet contamination coupling aging and electrical characteristic monitoring; contact angle, self-cleaning and interface strength tests are carried out simultaneously on flat or small-sized curved surface samples prepared in the same batch, thereby ensuring the repeatability of data and the feasibility of equipment integration.

[0080] To illustrate the relationship between the data acquisition process and subsequent analysis, this embodiment provides a set of sample data (10 samples, total aging time 2000h, sampling interval 24h) that can be collected and calculated by this system. The sample data is only used to illustrate the algorithm flow and table fields and does not constitute a limitation on the actual test results. The following table, using sample S01 as an example, lists some of the original data records of the accelerated aging process, as shown in Table 1.

[0081] Table 1. Raw data record of accelerated aging of sample S01

[0082]

[0083] The original data shown in Table 1 was used to generate Figure 3 The degradation curves of key functional parameters are used as the input basis for data preprocessing, feature extraction and lifetime prediction in steps S3 to S5.

[0084] As a preferred embodiment, the degradation criterion in this embodiment is as follows: when any of the following conditions are met, the sample is labeled as having superhydrophobic functional degradation for model training and verification: ① Static contact angle Or contact angle hysteresis ② Roll angle ③ Self-cleaning efficiency ④ Leakage current under specified voltage and wet / polluted conditions Exceeding the threshold Or it may exhibit characteristics of persistent partial discharge.

[0085] To improve the engineering applicability of the criteria, the threshold can be set as a grade: when the static contact angle is less than 150° or the roll-off angle is greater than 10° or the self-cleaning efficiency is less than 0.80, it is judged as "super-dual hydrophobic function attenuation"; when the static contact angle is less than 120°, the roll-off angle is greater than 25° or the self-cleaning efficiency is less than 0.60, it is judged as "functional failure". Electrical characteristics such as leakage current and partial discharge can be combined as mandatory failure calibration conditions.

[0086] The data then proceeds to step S3, the data fusion preprocessing stage. First, an improved method is used... Criteria combined with sliding window mid-range filtering to remove outliers: for any parameter sequence If a point deviates from the local mean by more than three times the dynamic standard deviation, it is considered an anomaly and removed; secondly, time alignment is performed, based on the aging start time. Based on this, linear interpolation is performed on six types of asynchronously sampled data to align them to a unified time grid. Next, a Savitzky-Golay filter with a window length of 5 and a second-order polynomial is applied to smooth each type of data to suppress high-frequency noise. Furthermore, for parameters with monotonically decaying characteristics such as contact angle and roll-off angle, an exponential decay model is adopted. The long-term trend term is fitted and then subtracted from the original sequence to retain short-term fluctuation information. Finally, cross-modal normalization is performed. First, Min-Max normalization is applied to the six different dimensional data to the [0,1] interval, and then a nonlinear weighting factor is introduced. Compression adjustment was performed, among which Indicates the first The current rate of change of the class parameter, and An empirical constant determined based on historical data statistics (usually taken as...). , ).

[0087] After outlier removal, time alignment, smoothing, and normalization, a unified data format for cross-modal fusion modeling is obtained. Taking sample S01 as an example, the data excerpts after normalization of the six core indicators are shown in Table 2.

[0088] Table 2. Fusion data of sample S01 after normalization

[0089]

[0090] The normalized data in Table 2 correspond one-to-one with the original data in Table 1, which facilitates inputting multi-source data with different dimensions and different sampling noise levels into the same feature engineering and prediction model.

[0091] Step S4: Construct an 8-dimensional degenerate state vector The process includes feature extraction at five levels: regarding macroscopic performance degradation features, calculating the contact angle decay rate. Roll-off angle growth rate Self-cleaning efficiency decline rate Regarding microstructure degradation characteristics, the integrity index of micro / nano structures is calculated based on laser confocal imaging. ,in For the initial roughness, For roughness after aging; in terms of chemical composition degradation characteristics, a low surface energy migration index is defined. This reflects the degree of functional component loss; regarding the characteristics of mechanical interface degradation, the slope of the interface bonding strength attenuation is extracted. Regarding the characteristics of time-series cumulative damage, a cumulative aging factor was constructed. The weight Principal component analysis (PCA) automatically determines the principal components from the training samples. The selection criteria for PCA principal components are: cumulative variance contribution rate ≥ 95%, ensuring that the selected principal components can retain the core information of the original data. First, a candidate feature set of no less than 12 dimensions is constructed (based on contact angle, roll-off angle, self-cleaning efficiency, micro / nano morphology, component migration, and interfacial bonding strength, further introducing contact angle hysteresis). Pollution coverage Leakage current Candidate features (such as optional electrical features) are selected and Z-score standardized. Then, based on mutual information and recursive feature elimination, the eight features with the highest information gain are selected to form the final degenerate state vector. .

[0092] Furthermore, based on mutual information feature selection, this embodiment selects eight candidate features with high information gain to form a degradation state vector, and then normalizes it using Z-score. Taking sample S01 as an example, the values ​​of its 8-dimensional degradation state vector at different aging times are shown in Table 3.

[0093] Table 3. Values ​​of the eight-dimensional degenerate state vector of sample S01

[0094]

[0095] The 8-dimensional degradation state vector in Table 3 is used as the input to the LSTM-Attention model in step S5; among them, degradation rate features, cumulative aging factor, and instantaneous contact angle / roll angle jointly characterize the temporal degradation state.

[0096] Step S5 will degenerate the state vector Input the improved LSTM-Attention lifetime prediction model. For example... Figure 6 As shown, the model adopts an architecture of "8 input layers - two-layer LSTM (64 units per layer) - Attention weight layer - single output layer". The core calculation formula of the Attention mechanism is as follows:

[0097]

[0098] In the formula, For LSTM at time The hidden state, For learnable parameters, This is a context vector used for weighted fusion of key degradation information from each time step. The output layer uses a linear activation function to predict the remaining lifetime. The model training uses the Huber loss function to ensure robustness against outliers, while an early stopping mechanism is introduced to prevent overfitting.

[0099] The effectiveness of the present invention is verified through more specific test examples below.

[0100] Typical climatic characteristics (high humidity, frequent fog, heavy rainfall, and organic aerosols from vegetation / tea gardens, etc.) of the power line corridor in Wuyishan area, Fujian Province, were selected for comparative verification under two paths: natural exposure and artificial acceleration. First, the same batch of ultra-dual-hydrophobic coated insulators were deployed at field exposure points in Wuyishan, and contact angle, roll-off angle, self-cleaning efficiency, pollution coverage, and electrical characteristics were periodically collected. Second, based on the statistical characteristics of the field environment, a cyclic profile of "humid heat-condensation-ultraviolet light-pollution-electric field" was set in a simulation chamber to achieve equivalent acceleration. The same fusion preprocessing and feature construction process was used for the data from both paths to obtain degradation state vectors. The data were then input into the lifetime model, and the consistency between the predicted Remaining Useful Life (RUL) and the field-calibrated RUL was compared. The same fusion preprocessing and feature construction process was used on the data from both paths to obtain the degradation state vector F, which was then input into the lifetime model. The results showed that the correlation coefficient between the artificially accelerated aging predicted RUL and the naturally exposed calibrated RUL was 0.94, with an average relative error of 4.8%, verifying the prediction accuracy of the method in complex service environments. Figures 3 to 6 Example graphs of key performance degradation curves, feature correlation / dimensionality reduction results, lifetime prediction comparisons, and risk level evolution are provided to support the feasibility study in the Wuyishan scenario.

[0101] in, Figure 4 Among them, (a) is the correlation matrix of degradation features, which intuitively shows the linear correlation between each feature (such as contact angle decay rate and micro / nano structure integrity index) in the 8-dimensional degradation state vector. The color depth corresponds to the magnitude of the correlation coefficient; (b) is the PCA dimensionality reduction projection map (color represents aging time), which projects the 8-dimensional high-dimensional features to a two-dimensional space, clearly showing the feature clustering effect of different aging stages. The longer the aging time, the more concentrated the projection points are in a specific area.

[0102] Figure 5 Among them, (a) is a comparison chart of predicted and actual values, with the horizontal axis representing aging time and the vertical axis representing remaining service life. The solid line represents the actual value and the dashed line represents the predicted value. The two trends are highly consistent. (b) is a prediction error distribution chart. The normal distribution is used for fitting, and the error is concentrated within ±5%, indicating that the model prediction accuracy is high.

[0103] Figure 6 Among them, (a) is the evolution diagram of failure risk probability over time, showing the changing trend of the probability proportion of low, medium and high risks with aging time, and the probability of high risk gradually increases in the later stage of aging; (b) is the decision logic diagram of risk level, which uses the comparison of remaining useful life (RUL) with set thresholds T1 and T2 and risk level as the judgment conditions to output the decision path of health status, early warning signal and failure diagnosis report in sequence, and clarifies the judgment criteria of each node.

[0104] After the model outputs results, a decision is made based on preset logic: if the prediction... (For example If the risk level is low (year) and the insulator is considered to be in good health; if (For example If the risk level is medium (year) or higher, an early warning will be triggered and a shorter on-site testing cycle will be recommended; if If the risk level is high, then further analysis of the degradation state vector is needed. The dominant degradation pattern in, for example, when When the failure rate reaches a certain threshold, it indicates that the micro / nano structure has significantly collapsed. At this point, the system automatically generates a failure mechanism diagnosis report and recommends corresponding maintenance strategies (such as recoating with a superhydrophobic coating or replacing the insulator). Furthermore, if the prediction uncertainty (estimated by the Monte Carlo Dropout method, where hidden units in the LSTM layer are randomly deactivated with a 50% probability during the model inference phase, and the inference is repeated 20 times, with the standard deviation of the 20 predictions being calculated, which is the prediction uncertainty) exceeds 15%, the system will automatically initiate a secondary accelerated aging experiment. After collecting 5 new sets of data, the local model will be retrained and the prediction results updated to ensure evaluation accuracy.

[0105] Throughout the implementation process, the system can also achieve automated operation through electronic devices. These electronic devices include a processor, a memory, and a bus. The processor and memory are connected via the bus. The memory stores a set of program code, and the processor calls this program code to execute all the aforementioned steps. Similarly, a non-volatile computer storage medium can also store computer-executable instructions for executing the weather resistance assessment and lifetime prediction system for the super-double-hydrophobic insulators of this invention. In summary, this invention, by constructing a multi-physics field coupled accelerated aging environment, deploying a high-precision in-situ characterization platform, fusing multi-source heterogeneous sensor data, extracting multi-scale degradation features, and combining a deep learning model for lifetime prediction, achieves a systematic assessment and intelligent lifetime management of the weather resistance performance of super-double-hydrophobic insulators under complex service conditions. This effectively solves the problem in existing technologies of lacking comprehensive evaluation methods for the stability of micro / nano structures, surface functional degradation mechanisms, and lifetime evolution laws under multi-stress coupling.

[0106] Based on the same inventive concept, this invention also provides a computer device, comprising: one or more processors, and a memory for storing one or more computer programs; the programs include program instructions, and the processor executes the program instructions stored in the memory. The processor may be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. It is the computing and control core of the terminal, used to implement one or more instructions, specifically for loading and executing one or more instructions stored in a computer storage medium to implement the above-described method.

[0107] It should be further explained that, based on the same inventive concept, the present invention also provides a computer storage medium storing a computer program, which, when executed by a processor, performs the above-described method. This storage medium can be any combination of one or more computer-readable media. A computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. A computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of computer-readable storage media (a non-exhaustive list) include: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In the present invention, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0108] It should be noted that, unless otherwise defined, the technical or scientific terms used in this invention should have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0109] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

[0110] This invention is not limited to the preferred embodiment described above. Anyone inspired by this invention can derive other various forms of weather resistance assessment and life prediction methods and systems for super-dual-diffuse insulators. All equivalent variations and modifications made within the scope of the claims of this invention shall fall within the scope of this invention.

Claims

1. A method for assessing the weather resistance and predicting the lifespan of a super-damp insulator, characterized in that, include: Applying temperature and humidity, ultraviolet radiation, dirt deposition, and multi-physics coupling of electric fields to super-dual-hydrophobic insulators accelerates aging. Multi-source time-series degradation data were collected simultaneously during the aging process. The data covered parameters such as surface hydrophobicity, micro-nano morphology, low surface energy component mobility, and coating interface bonding strength. Based on the multi-source temporal degradation data, a degradation state vector that integrates multi-scale degradation features is extracted. The features include at least an integrity index reflecting the relative change in micro / nano structure coverage and roughness, a migration index based on the time-varying ratio of low surface energy component concentration, and a temporal damage factor that is a weighted accumulation of the change rates of multiple degradation parameters. The degradation state vector is input into a pre-trained temporal deep learning model, which outputs the remaining service life and failure risk level of the super double-diffuse insulator.

2. The method for weather resistance assessment and life prediction of a super-diffuse insulator according to claim 1, characterized in that: The parameters for the multiphysics-coupled accelerated aging are set in conjunction with the target service environment and acceleration factor. The core parameter ranges include: a temperature and humidity cycling range of -10℃ to 40℃, a relative humidity of 30% to 95%, and an ultraviolet irradiation intensity of 0.63 to 1.20 W / m² at a narrow band control point of 340 nm. 2 / nm, with a fouling settling rate of 0.1 to 0.5 mg / cm³. 2 • day, applied voltage is ±50kV.

3. The method for weather resistance assessment and life prediction of a super-diffuse insulator according to claim 1, characterized in that: The integrity index is the product of the micro / nano structure coverage and the attenuation coefficient of the root mean square roughness after aging relative to the initial root mean square roughness; the migration index is the ratio of the real-time migration rate of low surface energy components during aging to the initial migration rate. The weighting coefficients of the temporal damage factor are automatically determined from the training samples through principal component analysis.

4. The method for weather resistance assessment and life prediction of a super-diffuse insulator according to claim 1, characterized in that: Before extracting multi-scale degradation features, multi-source time-series degradation data are fused and preprocessed. The preprocessing includes time alignment, data smoothing, trend correction, and cross-modal normalization. The cross-modal normalization introduces a nonlinear weight factor based on the parameter change rate for adjustment.

5. The method for weather resistance assessment and life prediction of a super-diffuse insulator according to claim 1, characterized in that: The temporal deep learning model is a long short-term memory network model that includes an attention mechanism.

6. The method for weather resistance assessment and life prediction of a super-diffuse insulator according to claim 1, characterized in that: Data acquisition employs a strategy of combining master samples with control samples from the same batch: master samples are used for monitoring electric field and wet-contamination coupling aging and electrical characteristics, while control samples from the same batch are used for testing contact angle, self-cleaning efficiency, and interfacial bonding strength.

7. The method for weather resistance assessment and life prediction of a super-diffuse insulator according to claim 1, characterized in that: After outputting the remaining service life and failure risk level, the following decision-making logic applies: If the remaining service life is not less than the first set threshold and the failure risk level is low, it is judged to be in a healthy state; if the remaining service life is not less than the second set threshold and less than the first set threshold, or the failure risk level is medium, an early warning is triggered; if the remaining service life is less than the second set threshold or the failure risk level is high, a failure mechanism diagnosis report and maintenance strategy are generated in combination with the dominant degradation mode; if the prediction uncertainty exceeds the preset threshold, a secondary accelerated aging experiment is automatically started, the local model is retrained after supplementing the data, and the prediction results are updated.

8. The method for weather resistance assessment and life prediction of a super-diffuse insulator according to claim 1, characterized in that: In the multi-source temporal degradation data: The migration rate of the low surface energy components is obtained by periodically detecting the change rate of fluorine or silicon surface concentration using a surface chemical analysis module. The coating interface bonding strength is obtained by applying a vertical peeling force using a micro-force sensor and recording the critical desorption load. All data is stored using a unified timestamp.

9. The method for weather resistance assessment and life prediction of a super-diffuse insulator according to claim 5, characterized in that: When training the time-series deep learning model, the Huber loss function is used as the optimization objective, and an early stopping mechanism is introduced to prevent overfitting.

10. A system for assessing the weather resistance and predicting the lifespan of ultra-damp insulators, characterized in that, For performing the method according to any one of claims 1 to 9, comprising: The multiphysics environment simulation chamber is configured to simultaneously apply temperature and humidity stress, ultraviolet radiation stress, pollution settlement stress and electric field stress to the super-dual-hydrophobic insulator. The in-situ characterization platform is configured to simultaneously acquire multi-source time-series degradation data during the aging process, including a high-speed camera system, a laser confocal microscope, a surface chemical analysis module, a micro-force sensor, and a six-axis robotic arm. The surface chemical analysis module is equipped with a sealed sampling and vacuum transfer component to achieve quasi-in-situ analysis. The data processing unit is communicatively connected to the in-situ characterization platform and is configured to perform data fusion preprocessing, multi-scale degradation feature extraction, degradation state vector construction, and lifetime prediction model calculation.

Images