Silica gel line bending attenuation prediction method and system based on impedance fluctuation spectrum

Abstract

The application provides a silicon rubber wire bending attenuation prediction method and system based on impedance fluctuation spectrum. It belongs to the technical field of electronic material reliability, electrochemical impedance spectrum analysis and intelligent life prediction. The method comprises the following steps: installing a silicon rubber wire on a four-probe electrochemical workstation test tool, setting cycle bending parameters and starting the test, and collecting impedance fluctuation spectrum data of the silicon rubber wire in the cycle bending process in real time; positioning the conductor and insulating layer interface region based on the impedance fluctuation spectrum data, and generating interface impedance time series data set; and monitoring the impedance fluctuation spectrum of the silicon rubber wire in the cycle bending process through the four-probe electrochemical workstation, which can accurately capture the electrochemical degradation process of the conductor and insulating layer interface, greatly improve the perception ability of micro damage, and make early failure warning possible.

Description

Technical Field

[0001] This invention proposes a method and system for predicting the tortuosity attenuation of silicone wires based on impedance fluctuation spectrum, which belongs to the interdisciplinary field of electronic material reliability, electrochemical impedance spectroscopy analysis and intelligent lifetime prediction. Background Technology

[0002] With the increasing integration and flexibility of electronic devices, silicone wires, as critical connection components, directly impact equipment reliability due to their bending resistance. Traditional evaluation techniques have significant limitations: tensile testing machines and similar equipment can only perform single-cycle tensile tests to fracture, failing to reproduce the cumulative damage process caused by cyclic bending stress in actual operating conditions, leading to test results that are out of sync with real-world service environments. Existing evaluation systems rely excessively on macroscopic parameters such as tensile strength and elongation, neglecting the electrochemical degradation that occurs at the conductor-insulator interface during bending. This microscopic damage can cause abnormally high contact resistance, ultimately leading to connection failure.

[0003] Lifetime prediction methods generally suffer from lag, requiring the sample to completely fracture before data can be obtained, thus failing to provide early fault warnings. More importantly, existing technologies have not established a quantitative relationship between bending strain and internal crack propagation and resistance changes; the prediction models are essentially black-box systems, lacking physical mechanism support. Although some patents mention flexible cable bending tests or online resistance measurements, none have overcome the limitations of functional silos. They have not deeply integrated impedance fluctuation spectrum analysis, multi-scale physical modeling, and deep learning feature extraction, making it difficult to construct a cognitive-level lifetime prediction architecture with non-destructive perception, physical mechanism modeling, early warning, and accelerated verification capabilities. This fails to meet the urgent needs of next-generation electronic materials for intelligent reliability assessment. Summary of the Invention

[0004] This invention provides a method and system for predicting the tortuosity attenuation of silicone wires based on impedance fluctuation spectrum, in order to solve the problems mentioned in the background art above:

[0005] The present invention proposes a method for predicting the tortuosity attenuation of silicone wires based on impedance fluctuation spectrum, the method comprising:

[0006] S1. Install the silicone wire on the four-probe electrochemical workstation test fixture, set the cyclic bending parameters and start the test, and collect the impedance fluctuation spectrum data of the silicone wire in real time during the cyclic bending process; locate the interface region between the conductor and the insulation layer based on the impedance fluctuation spectrum data, and generate the interface impedance time series dataset.

[0007] S2. Input the interface impedance time series dataset into a deep separable convolutional neural network to perform multi-scale feature extraction and dimensionality reduction to generate an impedance feature vector matrix; simultaneously construct a physical field coupled dynamic model to map the impedance feature vector matrix to the stress and electrochemical coupled fields, and calculate the crack propagation rate distribution cloud map.

[0008] S3. Damage density evolution analysis is performed using crack propagation rate distribution cloud map to generate three-dimensional damage density field data; a correlation model between resistance change and crack density is established based on the three-dimensional damage density field data, and the real-time resistance degradation coefficient is obtained through the model output.

[0009] S4. Input the real-time resistance degradation coefficient into the pre-trained lifetime prediction neural network. This network integrates impedance fluctuation spectrum characteristics with physical field coupling parameters and outputs the remaining lifetime prediction value of the silicone wire. Simultaneously generate a damage heat map to mark the distribution location of high-risk areas along the axial direction of the silicone wire.

[0010] S5. Perform a weighted confidence assessment on the remaining lifetime prediction value to generate a multi-dimensional reliability index; when the reliability index is lower than the preset threshold, trigger a graded early warning mechanism to generate early warning data on the bending attenuation of the silicone wire.

[0011] The present invention proposes a system for predicting the tortuosity attenuation of silicone wires based on impedance fluctuation spectrum, comprising:

[0012] One or more processors;

[0013] Memory, used to store one or more programs;

[0014] Wherein, when the one or more programs are executed by the one or more processors, the one or more processors are made to implement the method described in any one of the above.

[0015] The beneficial effects of this invention are as follows: By using a four-probe electrochemical workstation to monitor the impedance fluctuation spectrum of silicone wires during cyclic bending in real time, the electrochemical degradation process at the conductor-insulator interface can be accurately captured, significantly improving the ability to detect microscopic damage and enabling early fault warning. Combining a deep separable convolutional neural network with a physical field coupled dynamics model, it achieves accurate prediction of key parameters such as crack propagation rate and resistance degradation coefficient, with prediction errors controlled within 10%, while also enhancing the interpretability of the prediction results and breaking the limitations of traditional black-box models. This method can complete the test without destroying the sample, reducing material waste and testing costs, while avoiding the evaluation lag caused by single tensile fracture tests, preventing sudden equipment failures due to misjudgment of lifespan. It not only meets the stringent reliability requirements of the flexible development of electronic devices for connecting components, but also provides data support for the R&D optimization of silicone wires, helping companies shorten product iteration cycles and enhance market competitiveness. Attached Figure Description

[0016] Figure 1 This is a diagram illustrating the steps of the method described in this invention. Detailed Implementation

[0017] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention. Example 1

[0018] One embodiment of the present invention, such as Figure 1 As shown, a method for predicting the tortuosity attenuation of silicone wires based on impedance fluctuation spectrum is described, the method comprising:

[0019] S1. Install the silicone wire on the four-probe electrochemical workstation test fixture, set the cyclic bending parameters and start the test, and collect the impedance fluctuation spectrum data of the silicone wire in real time during the cyclic bending process; locate the interface region between the conductor and the insulation layer based on the impedance fluctuation spectrum data, and generate the interface impedance time series dataset.

[0020] S2. Input the interface impedance time series dataset into a deep separable convolutional neural network to perform multi-scale feature extraction and dimensionality reduction to generate an impedance feature vector matrix; simultaneously construct a physical field coupled dynamic model to map the impedance feature vector matrix to the stress and electrochemical coupled fields, and calculate the crack propagation rate distribution cloud map.

[0021] S3. Damage density evolution analysis is performed using crack propagation rate distribution cloud map to generate three-dimensional damage density field data; a correlation model between resistance change and crack density is established based on the three-dimensional damage density field data, and the real-time resistance degradation coefficient is obtained through the model output.

[0022] S4. Input the real-time resistance degradation coefficient into the pre-trained lifetime prediction neural network. This network integrates impedance fluctuation spectrum characteristics with physical field coupling parameters and outputs the remaining lifetime prediction value of the silicone wire. Simultaneously generate a damage heat map to mark the distribution location of high-risk areas along the axial direction of the silicone wire.

[0023] S5. Perform a weighted confidence assessment on the remaining life prediction value to generate a multi-dimensional reliability index; when the reliability index is lower than a preset threshold, trigger a graded early warning mechanism to generate early warning data on the bending resistance attenuation of the silicone wire, which includes failure probability, risk level and maintenance recommendations.

[0024] The working principle and effects of the above technical solution are as follows: By capturing the interface impedance changes during the bending process of the silicone wire in real time through impedance fluctuation spectrum, a precise interface impedance time-series dataset is generated, improving the accuracy of damage feature extraction and reducing interference from invalid data. Combining a deep separable convolutional neural network with a physical field coupled dynamics model amplifies the correlation between damage and impedance signals, enhancing the accuracy of crack propagation rate calculation and reducing the error of traditional prediction methods. Through three-dimensional damage density field analysis and resistance degradation coefficient derivation, the one-sidedness of single-parameter prediction is avoided, reducing the deviation in lifetime prediction and preventing premature failure or excessive repair of the silicone wire due to inaccurate prediction. It can accurately output the remaining lifetime prediction value and identify high-risk areas, enhancing the practicality of the prediction. Multi-dimensional reliability assessment and graded early warning improve the timeliness of early warning, avoiding safety hazards caused by undetected degradation of the silicone wire during use, while reducing maintenance costs and balancing prediction accuracy with practical application needs.

[0025] In one embodiment of the present invention, S1 includes:

[0026] S11. Select silicone wire samples from the same batch and complete surface cleaning treatment. Accurately assemble the samples into the test fixture of the four-probe electrochemical workstation, adjust the contact posture between the probe and the silicone wire conductor and lock the fixed structure to ensure stable and reliable electrical connection of the test circuit.

[0027] S12. Input the operating parameters into the test control system. The operating parameters include the cyclic bending amplitude frequency and the reciprocating cycle, etc., calibrate the bending mechanism's motion stroke and start / stop nodes, start the continuous reciprocating bending loading program, and construct a dynamic test environment that fits the actual working conditions.

[0028] S13. During the entire cycle of cyclic bending, the high-frequency synchronous acquisition mode is activated to continuously capture the electrical response signal at the interface between the silicone wire conductor and the insulation layer, and to fully record the impedance amplitude and phase change information under different bending stages, forming the original impedance fluctuation spectrum data sequence.

[0029] S14. Perform time-domain and frequency-domain joint noise reduction on the original impedance fluctuation spectrum data to remove abnormal values ​​introduced by environmental electromagnetic interference and mechanical vibration, complete data baseline calibration and trend smoothing, and improve the overall effectiveness and continuity of the data.

[0030] S15. Based on the impedance fluctuation spectrum data after noise reduction, track the response change law of the interface region, split the impedance characteristic information of different time nodes, organize and summarize various effective data according to the time series, and finally generate a standardized interface impedance time series dataset.

[0031] The working principle and effects of the above technical solution are as follows: By standardizing the silicone wire sample processing and tooling assembly process, the stability of the test circuit connection is ensured, reducing test errors caused by poor contact and improving the reliability of the test foundation. Accurate input and calibration of bending operation parameters create a test environment that closely matches actual working conditions, avoiding test data distortion caused by parameter deviations and enhancing the realism of the test scenario. High-frequency synchronous acquisition mode completely captures impedance signals at different bending stages, ensuring the integrity of the original data and reducing the omission of key features. Time-domain and frequency-domain joint noise reduction and baseline calibration processing eliminate various interference factors, improving data validity and continuity, and avoiding interference from invalid data in subsequent analysis. By regularizing time-series features to generate a standardized dataset, both data format uniformity and the variation law of interface impedance are highlighted, providing high-quality data support for subsequent feature extraction and damage analysis, further improving the accuracy of the entire prediction process and reducing potential errors in subsequent steps.

[0032] In one embodiment of the present invention, S15 includes:

[0033] Extract the noise-reduced impedance fluctuation spectrum data, focus on tracking the impedance response change law of the interface region between the silicone wire conductor and the insulation layer, and capture the impedance value fluctuation difference under different bending periods.

[0034] Impedance characteristic information is broken down according to the test timing nodes, and the core characteristics corresponding to each node are separated. The core characteristics include impedance amplitude, phase and rate of change, etc., to complete the preliminary classification and organization of characteristic information.

[0035] The features of each time-series node after classification are numerically normalized to unify the data format and units, and invalid feature records with disordered time series and abnormal values ​​are removed, while valid data that conforms to the test rules are retained.

[0036] The normalized effective features are summarized in the order of test acquisition to form a continuous and complete time series feature sequence, and finally a standardized interface impedance time series dataset is generated.

[0037] The working principle and effects of the above technical solution are as follows: By focusing on tracking the impedance changes at the interface between the silicone wire conductor and the insulation layer, the impedance fluctuation differences at different bending periods are accurately captured, improving the targeting of interface feature extraction and reducing interference from irrelevant features. Core features are split and classified according to time-series nodes, clearly distinguishing the impedance amplitude, phase, and rate of change of each node, avoiding analytical biases caused by feature confusion, and enhancing the systematic nature of feature information. Time-series features are numerically normalized and invalid records are removed, unifying data format and units, improving data standardization and purity, and preventing abnormal data from affecting subsequent processing results. Effective features are summarized in the order of collection to form a continuous time-series sequence, generating a standardized dataset. This ensures both the continuity and integrity of the data, highlights the temporal pattern of impedance changes, provides high-quality data support for subsequent model input, further reduces errors in subsequent steps, and improves the stability of the entire prediction process.

[0038] In one embodiment of the present invention, S2 includes:

[0039] S21. Standardize and convert the interface impedance time series dataset, divide it into training and validation subsets and expand the channel dimension to adapt to the input rules of the deep separable convolutional neural network, thereby reducing feature loss during data transmission and computation.

[0040] S22. In a deep separable convolutional neural network, multi-layer and multi-scale convolutional kernel operations are enabled to extract micro-fluctuation features, macro-trend features and periodic mutation features respectively, and to mine deep correlation information related to structural damage in impedance signals.

[0041] S23. Perform layer-by-layer dimensionality reduction and adaptive weight allocation on the feature information extracted from multiple scales, remove redundant feature components and retain core effective features, and generate a high-dimensional compact impedance feature vector matrix through global pooling and feature recombination operations.

[0042] S24. Combining the material properties and structural parameters of silicone wire, construct a coupled dynamic model of the interaction between stress field and electrochemical field, set boundary conditions and material constitutive relations, and improve the calculation rules related to field transfer and energy dissipation.

[0043] S25. Map the impedance eigenvector matrix to the constructed coupled field computational domain, simulate the internal micro-defect development process through iterative numerical calculation, quantify the crack propagation rate at different locations, and generate a global visualized crack propagation rate distribution cloud map.

[0044] The working principle and effects of the above technical solution are as follows: Standardization and dimensional expansion of the interface impedance time-series dataset are performed to adapt to neural network input rules, reducing feature loss, improving data utilization, and avoiding computational anomalies caused by data format incompatibility. Various features are extracted through multi-layer, multi-scale convolutional kernels to uncover the deep correlation between impedance signals and structural damage, enhancing the comprehensiveness of feature extraction and reducing the omission of key damage features. Dimensionality reduction and weight allocation are performed on multi-scale features, eliminating redundant components and generating a compact feature vector matrix, reducing computational load and improving computational efficiency. A coupled dynamic model is constructed by combining material properties, refining calculation rules, improving the realism of crack propagation simulation, and avoiding analytical errors caused by simulation bias. The feature matrix is ​​mapped to coupled field calculations to quantify crack propagation rates and generate visual cloud maps, accurately presenting the development state of internal micro-defects and intuitively displaying crack distribution differences, providing accurate basis for subsequent damage analysis, and further improving the scientificity and reliability of the entire prediction process.

[0045] In one embodiment of the present invention, step S22 includes:

[0046] In a deep separable convolutional neural network, multi-layer, multi-scale convolutional kernels are enabled. The kernel parameters are set according to different size gradients to cover the three dimensions of feature extraction: micro, meso, and macro. Feature mining operations are then initiated.

[0047] By using small-sized convolutional kernels to perform fine scanning of impedance time-series data, we can capture minute impedance fluctuation details, extract micro-fluctuation features corresponding to micro-damage of silicone wires, and form a set of micro-features.

[0048] By performing sliding operations on time-series data using medium-sized convolutional kernels, the overall trend of impedance signal changes with bending period is tracked, macroscopic trend features reflecting damage accumulation are extracted, and integrated to form a macroscopic feature sequence.

[0049] By capturing abrupt nodes in impedance signals using large-size convolutional kernels, periodic abrupt features corresponding to crack initiation and propagation are selected, and feature parameters corresponding to abrupt moments are marked.

[0050] By integrating micro-fluctuation characteristics, macro-trend characteristics, and periodic mutation characteristics, and linking the intrinsic relationship between each characteristic and the damage to the silicone wire structure, we can uncover the deep correlation information hidden in the impedance signal.

[0051] The working principle and effects of the above technical solution are as follows: By setting convolutional kernels with different size gradients, the three extraction dimensions of micro, meso, and macro are covered, making feature mining more comprehensive and avoiding feature bias caused by single-scale extraction, thus reducing the omission of key damage information. Small-sized convolutional kernels finely scan and capture minute impedance fluctuations, accurately extracting micro-damage features, improving the sensitivity of micro-damage identification, and preventing micro-defects from being ignored. Medium-sized convolutional kernels track the overall trend of impedance signals, extracting the macro-trend of damage accumulation, clearly presenting the damage development law, and enhancing the coherence of feature extraction. Large-sized convolutional kernels capture impedance abrupt change nodes, accurately screening crack initiation and expansion-related features, avoiding the masking of crack-related signals, and improving the accuracy of crack feature identification. By fusing the three types of features and associating them with the intrinsic relationship with structural damage, deep correlation information is mined, which can not only comprehensively capture various damage features, but also highlight the intrinsic logic of damage development, providing accurate support for subsequent feature processing and crack analysis, further reducing the error of subsequent calculations, and improving the reliability of the entire prediction process.

[0052] In one embodiment of the present invention, step S23 includes:

[0053] The extracted multi-scale feature information is summarized and integrated into three types of features: micro fluctuations, macro trends, and periodic mutations, forming a multi-dimensional feature set.

[0054] Perform layer-by-layer dimensionality reduction and compression operations on multi-dimensional feature sets, simplify feature dimensions through feature filtering algorithms, and reduce the computational resources occupied by invalid features;

[0055] The feature components after dimensionality reduction are adaptively weighted, strengthening the feature weights that are strongly correlated with damage to the silicone wire structure and weakening the redundant and irrelevant feature components.

[0056] Redundant features with a weight ratio below the threshold are removed, and core effective features that can accurately characterize the damage state are retained to complete feature purification.

[0057] A global pooling operation is performed on the purified core features to integrate the feature space distribution information. Then, the effective features of each dimension are integrated through feature recombination to generate a high-dimensional compact impedance feature vector matrix.

[0058] The working principle and effects of the above technical solution are as follows: By summarizing three types of multi-scale features to form a multi-dimensional set, comprehensive integration of feature information is achieved, avoiding the limitations of single feature analysis and reducing the omission of damage-related information. Layer-by-layer dimensionality reduction and compression simplifies feature dimensions, reducing the computational resources occupied by invalid features, lowering the subsequent computational load, improving overall computational efficiency, and avoiding waste of computational resources. Adaptive weight allocation is performed on the dimensionality-reduced features, strengthening the weight of damage-related features, weakening redundant components, and then eliminating low-weight invalid features, completing feature purification, improving the accuracy and effectiveness of features, and preventing redundant features from interfering with subsequent analysis. Global pooling and feature recombination integrate core features of each dimension to generate a high-dimensional compact feature vector matrix, which can retain key damage features while reducing data volume, providing high-quality support for subsequent coupled field mapping and crack analysis, and further improving the accuracy and stability of the entire prediction process.

[0059] In one embodiment of the present invention, S3 includes:

[0060] S31. Based on the crack propagation rate distribution cloud map, the damage generation and development process inside the silicone wire is gradually tracked according to the bending cycle sequence. The damage values ​​of each region at different times are statistically analyzed to depict the evolution path of damage with the increase of cycle number.

[0061] S32. Perform fine meshing along the axial, radial and circumferential directions of the silicone line, calculate the damage density value in a single mesh cell, analyze the damage gradient change between adjacent cells, and capture the evolution characteristics of local concentrated damage areas.

[0062] S33. Integrating temporal damage change information with spatial grid distribution data, and constructing three-dimensional damage density field data covering the entire silicone wire through three-dimensional interpolation and field quantity reconstruction calculations, thus fully presenting the spatial distribution of internal damage.

[0063] S34. Based on three-dimensional damage density field data, a quantitative correlation model between resistance change and crack density is built, the functional relationship between crack density change and resistance degradation is fitted, and a mathematical mapping rule for multivariable coupling is established.

[0064] S35. Substitute the real-time updated three-dimensional damage density field data into the correlation model to carry out forward extrapolation calculations, comprehensively consider the crack distribution morphology and propagation degree, and output the real-time resistance degradation coefficient corresponding to the current test moment.

[0065] The working principle and effects of the above technical solution are as follows: By using crack propagation rate distribution cloud maps as a basis, the damage development process is tracked sequentially, clearly depicting the damage evolution path, improving the coherence of damage development analysis, and avoiding the omission of key damage change nodes. Fine meshing is performed along the three-dimensional direction to accurately calculate the damage density of each element, analyze damage gradient changes, capture local concentrated damage characteristics, enhance the accuracy of damage localization, and reduce the occurrence of ignored local damage.

[0066] By integrating temporal and spatial data, a global damage density field is constructed through three-dimensional interpolation and field reconstruction, fully presenting the spatial distribution of internal damage and avoiding the one-sidedness of single-dimensional analysis. A quantitative correlation model between resistance change and crack density is built, fitting the functional relationship between the two, improving the scientific rigor of resistance degradation analysis and preventing inference errors caused by unclear correlation.

[0067] By substituting real-time damage data into the model and comprehensively considering crack distribution and propagation, a real-time resistance degradation coefficient is output. This not only accurately reflects the current damage state but also provides a reliable basis for subsequent lifetime prediction, further improving the accuracy of the entire prediction process and reducing potential errors in subsequent steps.

[0068] In one embodiment of the present invention, S32 includes:

[0069] A continuous mesh is generated along the axis of the silicone line, and uniformly distributed axial calculation units are arranged.

[0070] Layered meshing is carried out along the radial and circumferential directions of the silicone line to form a three-dimensional spatial mesh structure; the damage density value in the corresponding region is calculated by traversing each independent mesh cell.

[0071] By comparing the damage density difference between adjacent grid cells, the trend of damage gradient change is analyzed.

[0072] Track areas with abnormally high damage density and capture the evolutionary characteristics of locally concentrated damage areas.

[0073] The working principle and effects of the above technical solution are as follows: By dividing the silicone line into uniform grid units along its axial direction, the uniformity and consistency of axial damage analysis are ensured, avoiding deviations in axial damage distribution analysis and reducing the omission of local damage. Layered grid subdivision along the radial and circumferential directions forms a three-dimensional spatial grid structure, achieving full coverage of the silicone line without blind spots, enhancing the comprehensiveness of damage analysis and preventing damage areas in three-dimensional space from being ignored. Damage density is calculated by traversing each grid unit, accurately obtaining the damage degree of each region, improving the accuracy of damage values, and avoiding errors caused by overall estimation. The difference in damage density between adjacent grids is compared to analyze damage gradient changes, clearly presenting the damage propagation trend and reducing errors in judging damage development patterns. Tracking areas with abnormally high damage density captures the evolution characteristics of locally concentrated damage, enabling rapid location of high-risk damage sites and understanding the development trend of local damage, providing accurate local damage data support for subsequent three-dimensional damage density field construction and resistivity degradation analysis, further improving the reliability of the entire damage analysis process.

[0074] In one embodiment of the present invention, step S4 includes:

[0075] S41. The real-time resistance degradation coefficient is imported into the lifetime prediction neural network trained with a large number of samples, and the data is normalized and matched with the network input layer parameter specifications to ensure the smooth execution of subsequent calculation processes.

[0076] S42. The deep features extracted from impedance wave spectrum are fused with the parameter information obtained by physical field coupling calculation within the life prediction neural network. The correlation between damage and life is enhanced by feature splicing and weight fusion operations.

[0077] S43. Through multi-layer fully connected layer operation and nonlinear fitting of neural network, the remaining life numerical extrapolation is completed by integrating multi-source feature information, and the predicted value of the remaining life of the silicone wire under the current working condition is output.

[0078] S44. Based on the internal damage distribution data and the resistance degradation coefficient, perform spatial visualization rendering, use color gradient to distinguish areas with different damage levels, and generate an intuitive and clear global damage thermal map of the silicone line.

[0079] S45. Identify high-risk areas where damage values ​​exceed the critical range in the damage thermal map, record the coordinate information of the corresponding sections, and accurately determine the specific distribution location and length range of the high-risk areas along the axis of the silicone wire.

[0080] The working principle and effects of the above technical solution are as follows: Real-time resistance degradation coefficients are imported into the trained neural network to complete data normalization and parameter matching, ensuring smooth operation, avoiding interruptions caused by data incompatibility, and reducing operational failures. Deep features and coupled parameters are integrated within the network to strengthen the correlation between damage and lifetime, improve the correlation of lifetime prediction, and prevent bias caused by single-parameter prediction. Through multi-layer fully connected operations and nonlinear fitting, the remaining lifetime is inferred by comprehensively considering multi-source features, improving the accuracy of prediction values, reducing lifetime prediction errors, and avoiding misjudgments due to inaccurate predictions. A visual heatmap is generated based on damage data and resistance degradation coefficients, using color gradients to distinguish the degree of damage, enhancing the intuitiveness of damage distribution, and facilitating rapid understanding of the overall damage status. High-risk areas are identified in the heatmap and their locations and lengths are accurately marked, enabling rapid location of potential hazards and providing clear direction for maintenance, reducing safety risks caused by overlooking high-risk areas, while also reducing maintenance and troubleshooting costs, balancing prediction accuracy with actual maintenance needs.

[0081] In one embodiment of the present invention, S42 includes:

[0082] Extract the deep features formed by the impedance wave spectrum during network forward propagation and complete the uniform regularization of feature dimensions; retrieve various parameter information from the physical field coupling calculation output and complete the format alignment processing of the parameter sequence.

[0083] A feature concatenation operation is performed on the normalized deep features and the aligned parameter information to form a fused feature set;

[0084] Perform weight fusion operation on the fusion feature set to increase the transmission weight of key features in network computation;

[0085] Optimize the characterization effect of the fused features and amplify the intrinsic correlation between damage state and lifespan decay.

[0086] The working principle and effects of the above technical solution are as follows: By extracting deep features from the impedance wave spectrum and regularizing its dimensions, and retrieving and aligning the physical field coupling parameters, the inability to fuse features and parameters due to inconsistent specifications is avoided, reducing information loss during the fusion process. The regularized features and parameters are then concatenated to form a complete fused feature set, enhancing the comprehensiveness of feature information and preventing correlation bias caused by insufficient support from a single type of feature. Weighted fusion is performed on the fused features to increase the transmission weight of key features, weaken interference from irrelevant features, improve feature utilization efficiency, and prevent key damage information from being masked. The representation effect of the fused features is optimized, amplifying the intrinsic correlation between damage and lifetime decay, allowing the neural network to more accurately capture the correlation between the two, and reducing lifetime prediction bias. The entire process not only achieves effective fusion of multi-source information but also strengthens core correlations, providing high-quality feature support for subsequent remaining lifetime extrapolation, further improving the accuracy and stability of lifetime prediction, and avoiding prediction errors caused by improper feature fusion.

[0087] In one embodiment of the present invention, step S5 includes:

[0088] S51. From multiple dimensions such as data fitting accuracy, model generalization ability, and consistency of field quantity calculation, a weighted confidence assessment is carried out on the remaining lifetime prediction value, and scores for each dimension are calculated and assigned corresponding weight coefficients.

[0089] S52. Integrate multi-dimensional evaluation scores for comprehensive calculation to generate a multi-dimensional reliability index that can fully characterize the credibility of prediction results and intuitively reflect the stability of lifetime prediction results.

[0090] S53. Compare the calculated multi-dimensional reliability index with the preset critical threshold to determine whether the current silicone wire damage state has reached the warning trigger condition and distinguish between normal state and risk state.

[0091] S54. When the reliability index is lower than the preset threshold, the graded early warning mechanism is activated. The corresponding risk level is divided by combining the remaining life prediction value and the damage distribution range, and the failure probability value under the corresponding working condition is quantitatively calculated.

[0092] S55. Combining the failure probability risk level and the distribution location of high-risk areas, match the corresponding operation and maintenance solution to generate early warning data on the bending resistance attenuation of silicone wires. The early warning data on the bending resistance attenuation of silicone wires includes the failure probability risk level and maintenance suggestions.

[0093] The working principle and effects of the above technical solution are as follows: A weighted confidence assessment of the remaining lifetime prediction value is conducted from multiple dimensions, comprehensively considering various factors related to the prediction, improving the comprehensiveness of the assessment, avoiding the one-sidedness caused by single-dimensional assessment, and reducing misjudgments of prediction credibility. Integrating multi-dimensional assessment scores to generate a reliability index intuitively reflects the stability of the prediction results, making prediction credibility easier to judge and enhancing the practicality of the assessment results. Comparing the reliability index with critical thresholds accurately distinguishes between normal and risk states, improving the accuracy of early warning trigger judgment and avoiding false or missed early warnings. Initiating tiered early warnings and classifying risk levels and quantifying failure probabilities makes early warnings more targeted, reducing maintenance difficulties caused by ambiguous early warnings. Combining multiple information sources with maintenance plans generates early warning data containing key information, which can promptly remind users of the risk of silicone wire degradation and provide clear guidance for maintenance, avoiding maintenance delays caused by unclear early warnings, reducing safety hazards and maintenance costs, and balancing early warning accuracy with actual maintenance needs.

[0094] One embodiment of the present invention provides a system for predicting the tortuosity attenuation of silicone wires based on impedance fluctuation spectrum, comprising:

[0095] One or more processors;

[0096] Memory, used to store one or more programs;

[0097] Wherein, when the one or more programs are executed by the one or more processors, the one or more processors are made to implement the method described in any one of the above. Example 2

[0098] Example 1 can predict the bending attenuation of silicone wires based on impedance fluctuation spectrum. It can output the remaining lifetime prediction value and generate early warning data of silicone wire bending attenuation by coupling a deep separable convolutional neural network with a physical field dynamic model. However, in the actual cyclic bending test, the interface impedance time series data may still be mixed with abnormal responses caused by probe contact state perturbations, loading and unloading switching transients, mechanical rebound disturbances, and local short-term non-damaging fluctuations. Such abnormal responses may show amplitude abrupt changes or phase shifts similar to real interface damage in a single time series segment. If directly incorporated into the subsequent damage characterization link, it can easily lead to local abnormal amplification of the crack propagation rate distribution cloud map, thereby interfering with the three-dimensional damage density field data, resistance degradation coefficient, remaining lifetime prediction value, and graded early warning results.

[0099] To address the aforementioned technical problems, this embodiment 2 ensures that the data entering the lifetime prediction neural network and the hierarchical early warning link more accurately reflects the true interface damage state, reducing the interference of transient disturbances / pseudo-anomaly signals on subsequent results, including:

[0100] A1. Install the silicone wire on the four-probe electrochemical workstation test fixture, set the cyclic bending parameters and start the test, and collect the impedance fluctuation spectrum data of the silicone wire in real time during the cyclic bending process; locate the interface region between the conductor and the insulation layer based on the impedance fluctuation spectrum data, and generate the interface impedance time series dataset.

[0101] A2. Input the interface impedance time series dataset into a deep separable convolutional neural network to perform multi-scale feature extraction and dimensionality reduction to generate an impedance feature vector matrix; simultaneously construct a physical field coupled dynamic model, map the impedance feature vector matrix to the stress and electrochemical coupled fields, and calculate the crack propagation rate distribution cloud map.

[0102] A3. Damage density evolution analysis is performed using crack propagation rate distribution cloud maps to generate three-dimensional damage density field data; a correlation model between resistance change and crack density is established based on the three-dimensional damage density field data, and the real-time resistance degradation coefficient is obtained through the model output.

[0103] A4. Based on the interface impedance time series dataset, impedance eigenvector matrix, crack propagation rate distribution cloud map, and three-dimensional damage density field data, the candidate abnormal segments are periodically segmented and analyzed to extract the degree of rebound residual offset, the degree of periodic continuous enhancement, and the degree of field consistency, and to generate real interface damage characterization values.

[0104] Specifically, A4 includes:

[0105] A41. The interface impedance time series dataset is time-aligned according to a single bending cycle. Each bending cycle is divided into a loading stage, a peak stage, an unloading stage, and a springback stabilization stage. The impedance amplitude, phase, and rate of change information of the corresponding stage are extracted respectively.

[0106] A42. Based on the degree of impedance residual offset of the rebound stabilization stage relative to the initial stabilization stage of the same period, the rebound residual offset is obtained; wherein, the rebound residual offset is used to characterize whether the abnormal response still exists irreversibly after unloading and rebounding. The larger the rebound residual offset, the more likely the abnormal response corresponds to real interface damage rather than short-term disturbance.

[0107] A43. Based on the recurrence, growth continuity, and cumulative enhancement trend of candidate abnormal segments in multiple consecutive bending cycles, the periodic continuous enhancement amount is obtained; wherein, the periodic continuous enhancement amount is used to characterize whether the abnormal response has the characteristic of continuous existence and gradual enhancement across cycles. The larger the periodic continuous enhancement amount, the more the abnormal response conforms to the real damage cumulative evolution law.

[0108] A44. Combining crack propagation rate distribution cloud map and three-dimensional damage density field data, analyze the degree of matching between the temporal position of the candidate abnormal segment and the local high stress area, high crack propagation rate area and area with abnormally high damage density, to obtain the field quantity consistency degree; wherein, the field quantity consistency degree is used to characterize the consistency between impedance anomaly and physical field evolution results. The higher the field quantity consistency degree, the more likely the abnormal response is to originate from the actual structural degradation.

[0109] A45. A comprehensive weighted processing is performed on the rebound residual offset, the periodic sustained enhancement, and the field consistency. The rebound residual offset is assigned a first weight, the periodic sustained enhancement a second weight, and the field consistency a third weight. Under the combined effect of these three factors, a true interface damage characterization value for the corresponding candidate abnormal segment is generated. This true interface damage characterization value is used to comprehensively characterize the likelihood that the candidate abnormal segment originates from true interface damage.

[0110] The working principle and effects of the above technical solution are as follows: By decomposing candidate abnormal segments into different stages of a single bending cycle for segmented analysis, it is possible to distinguish between transient peak fluctuations and irreversible changes that persist after rebound; by introducing cross-cycle continuous enhancement features, it is possible to further identify the long-term evolution behavior corresponding to the accumulation of real damage; by introducing the consistency of field quantities, it is possible to verify the impedance anomalies with crack propagation and damage density evolution results, avoiding one-sided judgments based solely on single electrical signal fluctuations. Therefore, the relevance and reliability of identifying real interface damage can be significantly improved.

[0111] A5. For the same candidate anomalous segment, further extract the transient recovery degree, and combine the rebound residual offset and the periodic continuous enhancement amount to generate transient disturbance / pseudo-anomaly characterization values.

[0112] Specifically, A5 includes:

[0113] A51. Based on the impedance response changes during the loading, peak, unloading, and rebound stabilization phases, analyze whether the abnormal response decays rapidly after unloading, recovers rapidly during the rebound stabilization phase, and exhibits short-term burst characteristics between adjacent sampling points to obtain the transient recovery degree. The transient recovery degree is used to characterize whether the abnormal response has the characteristics of rapid recovery, short duration, and no irreversible residue. The higher the transient recovery degree, the more likely the abnormal response is to be a transient disturbance / pseudo-abnormal signal.

[0114] A52. Retrieve the rebound residual offset and periodic sustained enhancement amount obtained in step A4, and perform reverse constraint processing on the two to characterize the degree to which the candidate abnormal segment lacks irreversible residual features and cross-period sustained enhancement features.

[0115] A53. The inverse constraint results of transient recovery degree, rebound residual offset, and periodic sustained enhancement are comprehensively weighted. A first judgment weight is assigned to the transient recovery degree, a second judgment weight to the rebound residual offset, and a third judgment weight to the periodic sustained enhancement. Under the combined effect of these three factors, a transient disturbance / pseudo-anomaly characterization value for the corresponding candidate anomalous segment is generated. This transient disturbance / pseudo-anomaly characterization value is used to comprehensively characterize the likelihood that the candidate anomalous segment originates from a transient disturbance / pseudo-anomaly signal.

[0116] The working principle and effect of the above technical solution are as follows: By specifically introducing transient recovery degree, it is possible to separate abnormal segments that only appear briefly near the loading peak, recover rapidly after unloading and rebound, and do not have cross-cycle persistence from the real damage candidates; at the same time, by using the same rebound residual offset and cycle persistence enhancement amount from step A4 to participate in the reverse judgment, cross-entanglement and mutual verification between parameters can be formed, avoiding the situation where real damage is misjudged as a false anomaly based on a single recovery feature. Therefore, the recognition accuracy of transient disturbance / false anomaly signals can be improved, and the false alarm rate can be reduced.

[0117] A6. Based on the real interface damage characterization value and the transient disturbance / pseudo-anomaly characterization value, generate the credibility of the real interface damage, classify and screen candidate abnormal segments, and send the screened real interface damage signal into the subsequent lifetime prediction and hierarchical early warning link.

[0118] Specifically, A6 includes:

[0119] A61. Retrieve the true interface damage characterization value obtained in step A4 and the transient disturbance / pseudo-anomaly characterization value obtained in step A5, normalize and comprehensively compare the two, and generate the true interface damage credibility of the corresponding candidate anomaly segment. The true interface damage credibility is used to comprehensively reflect the credibility of the candidate anomaly segment belonging to the true interface damage signal. The higher the true interface damage characterization value and the lower the transient disturbance / pseudo-anomaly characterization value, the higher the true interface damage credibility.

[0120] A62. Compare the credibility of the actual interface damage with the preset credibility threshold. When the credibility of the actual interface damage is not lower than the preset credibility threshold, the corresponding abnormal segment is determined to be an actual interface damage signal. When the credibility of the actual interface damage is lower than the preset credibility threshold, the corresponding abnormal segment is determined to be a transient disturbance / pseudo-abnormal signal, and it is subjected to downweighting or elimination processing.

[0121] A63. The filtered real interface damage signals are re-aggregated to form a corrected interface damage feature sequence, and the real-time resistance degradation coefficient is corrected and updated based on the corrected interface damage feature sequence to generate a corrected remaining lifetime prediction value.

[0122] A64. Input the corrected remaining life prediction value into the subsequent weighted confidence assessment and graded early warning link to generate the corrected multi-dimensional reliability index, risk level, failure probability and maintenance suggestions, thereby obtaining silicone wire bending attenuation early warning data that is closer to the actual structural degradation state.

[0123] The working principle and effects of the above technical solution are as follows: By jointly constructing the credibility of true interface damage using real interface damage characterization values ​​and transient disturbance / pseudo-anomaly characterization values, the judgment method that relies solely on the magnitude of the anomaly amplitude can be transformed into a bidirectional competitive judgment method that simultaneously enhances true damage and suppresses pseudo-anomalies. By downweighting or eliminating low-credibility anomaly segments, the probability of transient disturbance signals entering the lifetime prediction link can be reduced. By retaining high-credibility anomaly segments and enhancing their role in subsequent links, the remaining lifetime prediction value and the silicone wire bending attenuation warning data can more realistically reflect the degradation state of the conductor and insulation layer interface. Therefore, it not only improves the differentiation accuracy between real interface damage signals and transient disturbance / pseudo-anomaly signals, but also further reduces the risk of false alarms and missed alarms in subsequent remaining lifetime prediction and graded warning.

[0124] The working principle and effect of this embodiment 2 are as follows: By specifically deepening the downstream identification logic, the original method of directly inputting the lifetime prediction neural network and weighted confidence assessment link based solely on the real-time resistance degradation coefficient is further refined into a processing method that includes characterization of real interface damage, characterization of transient disturbances / pseudo-anomalies, determination of the credibility of real interface damage, and correction of subsequent prediction links. This setup significantly improves the ability to identify real interface damage signals and reduces the interference of transient disturbances / pseudo-anomalies on crack propagation rate distribution cloud maps, three-dimensional damage density field data, real-time resistance degradation coefficients, remaining lifetime prediction values, and silicone wire flexural attenuation warning data.

[0125] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for predicting the tortuosity attenuation of silicone wires based on impedance fluctuation spectrum, characterized in that, The method includes: S1. Install the silicone wire on the four-probe electrochemical workstation test fixture, set the cyclic bending parameters and start the test, and collect the impedance fluctuation spectrum data of the silicone wire in real time during the cyclic bending process; locate the interface region between the conductor and the insulation layer based on the impedance fluctuation spectrum data, and generate the interface impedance time series dataset. S2. Input the interface impedance time series dataset into a deep separable convolutional neural network to perform multi-scale feature extraction and dimensionality reduction to generate an impedance feature vector matrix; simultaneously construct a physical field coupled dynamic model to map the impedance feature vector matrix to the stress and electrochemical coupled fields, and calculate the crack propagation rate distribution cloud map. S3. Damage density evolution analysis is performed using crack propagation rate distribution cloud map to generate three-dimensional damage density field data; a correlation model between resistance change and crack density is established based on the three-dimensional damage density field data, and the real-time resistance degradation coefficient is obtained through the model output. S4. Input the real-time resistance degradation coefficient into the pre-trained lifetime prediction neural network. This network integrates impedance fluctuation spectrum characteristics with physical field coupling parameters and outputs the remaining lifetime prediction value of the silicone wire. Simultaneously generate a damage heat map to mark the distribution location of high-risk areas along the axial direction of the silicone wire. S5. Weighted confidence assessment of the remaining life prediction value is performed to generate a multi-dimensional reliability index. When the reliability index is lower than the preset threshold, a graded early warning mechanism is triggered to generate early warning data on the bending resistance attenuation of silicone wire. In order to solve the problem of time-series alignment of the interface impedance time-series dataset according to a single bending cycle in actual bending process, based on the interface impedance time-series dataset, impedance feature vector matrix, crack propagation rate distribution cloud map and three-dimensional damage density field data, the candidate abnormal segments are periodically segmented and analyzed to extract the degree of springback residual offset, the degree of periodic continuous enhancement and the degree of field consistency, and generate real interface damage characterization values. Among them, the interface impedance time-series dataset is time-series aligned according to a single bending cycle, and each bending cycle is divided into loading stage, peak stage, unloading stage and springback stabilization stage.

2. The method for predicting the tortuosity attenuation of silicone wire based on impedance fluctuation spectrum according to claim 1, characterized in that, S1 includes: S11. Select silicone wire samples from the same batch and complete surface cleaning treatment. Accurately assemble the samples into the test fixture of the four-probe electrochemical workstation, adjust the contact posture between the probe and the silicone wire conductor, and lock and fix the structure. S12. Input the operating parameters into the test control system, calibrate the motion stroke and start / stop nodes of the bending mechanism, start the continuous reciprocating bending loading program, and build a dynamic test environment. S13. During the entire cycle of cyclic bending, the high-frequency synchronous acquisition mode is activated to continuously capture the electrical response signal at the interface between the silicone wire conductor and the insulation layer, and to fully record the impedance amplitude and phase change information under different bending stages, forming the original impedance fluctuation spectrum data sequence. S14. Perform time-domain and frequency-domain joint noise reduction on the original impedance fluctuation spectrum data to remove abnormal values ​​introduced by environmental electromagnetic interference and mechanical vibration, and complete data baseline calibration and trend smoothing. S15. Based on the impedance fluctuation spectrum data after noise reduction, track the response change law of the interface region, split the impedance characteristic information of different time nodes, organize and summarize various effective data according to the time series, and finally generate a standardized interface impedance time series dataset.

3. The method for predicting the tortuosity attenuation of silicone wire based on impedance fluctuation spectrum according to claim 1, characterized in that, The S2 includes: S21. Standardize and convert the interface impedance time series dataset, divide it into training and validation subsets, expand the channel dimension, and adapt it to the input rules of the depthwise separable convolutional neural network. S22. In a deep separable convolutional neural network, multi-layer and multi-scale convolutional kernel operations are enabled to extract micro-fluctuation features, macro-trend features and periodic mutation features respectively, and to mine deep correlation information related to structural damage in impedance signals. S23. Perform layer-by-layer dimensionality reduction and adaptive weight allocation on the feature information extracted from multiple scales, remove redundant feature components and retain core effective features, and generate a high-dimensional compact impedance feature vector matrix through global pooling and feature recombination operations. S24. Combining the material properties and structural parameters of silicone wire, construct a coupled dynamic model of the interaction between stress field and electrochemical field, set boundary conditions and material constitutive relations, and improve the calculation rules related to field transfer and energy dissipation. S25. Map the impedance eigenvector matrix to the constructed coupled field computational domain, simulate the internal micro-defect development process through iterative numerical calculation, quantify the crack propagation rate at different locations, and generate a global visualized crack propagation rate distribution cloud map.

4. The method for predicting the tortuosity attenuation of silicone wire based on impedance fluctuation spectrum according to claim 3, characterized in that, S22 includes: In a deep separable convolutional neural network, multi-layer, multi-scale convolutional kernels are enabled. The kernel parameters are set according to different size gradients to cover the three dimensions of feature extraction: micro, meso, and macro. Feature mining operations are then initiated. By using small-sized convolutional kernels to perform fine scanning of impedance time-series data, we can capture minute impedance fluctuation details, extract micro-fluctuation features corresponding to micro-damage of silicone wires, and form a set of micro-features. By performing sliding operations on time-series data using medium-sized convolutional kernels, the overall trend of impedance signal changes with bending period is tracked, macroscopic trend features reflecting damage accumulation are extracted, and integrated to form a macroscopic feature sequence. By capturing abrupt nodes in impedance signals using large-size convolutional kernels, periodic abrupt features corresponding to crack initiation and propagation are selected, and feature parameters corresponding to abrupt moments are marked. By integrating micro-fluctuation characteristics, macro-trend characteristics, and periodic mutation characteristics, and linking the intrinsic relationship between each characteristic and the damage to the silicone wire structure, we can uncover the deep correlation information hidden in the impedance signal.

5. The method for predicting the tortuosity attenuation of silicone wire based on impedance fluctuation spectrum according to claim 3, characterized in that, S23 includes: The extracted multi-scale feature information is summarized and integrated into three types of features: micro fluctuations, macro trends, and periodic mutations, forming a multi-dimensional feature set. Perform layer-by-layer dimensionality reduction and compression operations on multi-dimensional feature sets, simplify feature dimensions through feature filtering algorithms, and reduce the computational resources occupied by invalid features; The feature components after dimensionality reduction are adaptively weighted, strengthening the feature weights that are strongly correlated with damage to the silicone wire structure and weakening the redundant and irrelevant feature components. Redundant features with a weight ratio below the threshold are removed, and core effective features that can accurately characterize the damage state are retained to complete feature purification. A global pooling operation is performed on the purified core features to integrate the feature space distribution information. Then, the effective features of each dimension are integrated through feature recombination to generate a high-dimensional compact impedance feature vector matrix.

6. The method for predicting the tortuosity attenuation of silicone wire based on impedance fluctuation spectrum according to claim 1, characterized in that, The S3 includes: S31. Based on the crack propagation rate distribution cloud map, the damage generation and development process inside the silicone wire is gradually tracked according to the bending cycle sequence. The damage values ​​of each region at different times are statistically analyzed to depict the evolution path of damage with the increase of cycle number. S32. Perform fine meshing along the axial, radial and circumferential directions of the silicone line, calculate the damage density value in a single mesh cell, analyze the damage gradient change between adjacent cells, and capture the evolution characteristics of local concentrated damage areas. S33. Integrating temporal damage change information with spatial grid distribution data, and constructing three-dimensional damage density field data covering the entire silicone wire through three-dimensional interpolation and field quantity reconstruction calculations, thus fully presenting the spatial distribution of internal damage. S34. Based on three-dimensional damage density field data, a quantitative correlation model between resistance change and crack density is built, the functional relationship between crack density change and resistance degradation is fitted, and a mathematical mapping rule for multivariable coupling is established. S35. Substitute the real-time updated three-dimensional damage density field data into the correlation model to carry out forward extrapolation calculations, comprehensively consider the crack distribution morphology and propagation degree, and output the real-time resistance degradation coefficient corresponding to the current test moment.

7. The method for predicting the tortuosity attenuation of silicone wire based on impedance fluctuation spectrum according to claim 6, characterized in that, S32 includes: A continuous mesh is generated along the axis of the silicone line, and uniformly distributed axial calculation units are arranged. Layered meshing is carried out along the radial and circumferential directions of the silicone line to form a three-dimensional spatial mesh structure; the damage density value in the corresponding region is calculated by traversing each independent mesh cell. By comparing the damage density difference between adjacent grid cells, the trend of damage gradient change is analyzed. Track areas with abnormally high damage density and capture the evolutionary characteristics of locally concentrated damage areas.

8. The method for predicting the tortuosity attenuation of silicone wire based on impedance fluctuation spectrum according to claim 1, characterized in that, The S4 includes: S41. The real-time resistance degradation coefficient is imported into the lifetime prediction neural network trained with a large number of samples, and the data normalization process is completed and the network input layer parameter specifications are matched. S42. The deep features extracted from impedance wave spectrum are fused with the parameter information obtained by physical field coupling calculation within the life prediction neural network. The correlation between damage and life is enhanced by feature splicing and weight fusion operations. S43. Through multi-layer fully connected layer operation and nonlinear fitting of neural network, the remaining life numerical extrapolation is completed by integrating multi-source feature information, and the predicted value of the remaining life of the silicone wire under the current working condition is output. S44. Based on the internal damage distribution data and the resistance degradation coefficient, perform spatial visualization rendering, use color gradient to distinguish areas with different damage levels, and generate an intuitive and clear global damage thermal map of the silicone line. S45. Identify high-risk areas where damage values ​​exceed the critical range in the damage thermal map, record the coordinate information of the corresponding sections, and accurately determine the specific distribution location and length range of the high-risk areas along the axis of the silicone wire.

9. The method for predicting the tortuosity attenuation of silicone wire based on impedance fluctuation spectrum according to claim 1, characterized in that, The S5 includes: S51. Conduct a weighted confidence assessment of the remaining life prediction value from multiple dimensions, calculate the score of each dimension and assign corresponding weight coefficients; S52. Integrate multi-dimensional evaluation scores for comprehensive calculation to generate a multi-dimensional reliability index that can fully characterize the credibility of prediction results and intuitively reflect the stability of lifetime prediction results. S53. Compare the calculated multi-dimensional reliability index with the preset critical threshold to determine whether the current silicone wire damage state has reached the warning trigger condition and distinguish between normal state and risk state. S54. When the reliability index is lower than the preset threshold, the graded early warning mechanism is activated. The corresponding risk level is divided by combining the remaining life prediction value and the damage distribution range, and the failure probability value under the corresponding working condition is quantitatively calculated. S55. Combining the failure probability risk level and the distribution of high-risk areas, match the corresponding operation and maintenance solution to generate early warning data on the bending resistance and attenuation of silicone wires.

10. A system for predicting the tortuosity attenuation of silicone wires based on impedance fluctuation spectrum, characterized in that, The system includes: One or more processors; Memory, used to store one or more programs; Wherein, when the one or more programs are executed by the one or more processors, the one or more processors implement the method of any one of claims 1 to 9.

Images