Method and device for testing insulating glue

Abstract

The invention relates to the technical field of insulation paste testing, and discloses an insulation paste testing method and device, and the method comprises the steps: synchronously collecting the multi-dimensional performance parameters of to-be-tested insulation paste, extracting the feature vector of an initial data set, and carrying out the dynamic loading experiment of the to-be-tested insulation paste based on a performance feature matrix, constructing a time sequence curve of the dynamic response data set, performing multi-dimensional evaluation on the degradation parameter set based on information entropy to obtain a comprehensive score, and performing anomaly detection on the comprehensive score; by synchronously collecting the multi-dimensional performance parameters of the to-be-tested insulation paste, the performance characteristic matrix is constructed, so that the test data can comprehensively reflect the performance of the insulation paste in different dimensions; the comprehensive score is analyzed based on anomaly detection, potential quality hazards can be effectively identified, misjudgment caused by local index anomaly is avoided, quality grades are divided through iteration, it is ensured that a quality evaluation result can more reasonably guide model selection and application of the insulation paste, and the reliability of the insulation paste is improved.

Description

Technical Field

[0001] The invention relates to the technical field of insulating glue testing, and in particular to a method and a device for testing insulating glue. Background Art

[0002] Insulating adhesive is a key material widely used in electronics, electricity, aerospace and other fields. It is mainly used to provide electrical insulation, protective sealing and mechanical support. As electronic equipment develops towards high power and miniaturization, the performance requirements for insulating adhesives are also increasing. Not only do they need to have excellent electrical insulation properties, but they also need to have good mechanical strength, environmental stability and long-term reliability. For example, in high-voltage power transmission equipment, insulating adhesives need to withstand high electric field strengths while maintaining stable dielectric properties; in electronic packaging applications, insulating adhesives need to have excellent adhesion and resistance to thermal expansion to prevent cracks or peeling during long-term use.

[0003] In the prior art, the testing methods for insulating adhesives mostly adopt static measurement, that is, independently testing parameters such as resistivity and dielectric constant, and then making quality judgments through data comparison or empirical formulas. However, in the implementation process, it is usually difficult to fully capture the performance degradation characteristics of the material under dynamic use conditions, especially the performance under long-term or high-stress environments, resulting in the test results being unable to accurately reflect the long-term stability of the insulating adhesive in actual use.

[0004] Therefore, it is necessary to provide a testing method and device for insulating glue to solve the problem that the existing testing method for insulating glue cannot accurately reflect the long-term stability of the insulating glue in actual use. Summary of the invention

[0005] The main purpose of the present invention is to provide a testing method and device for insulating glue, aiming to solve the technical problems mentioned in the above background technology.

[0006] The present invention adopts the following technical solutions: A testing method for insulating adhesive, comprising: Synchronously collecting multi-dimensional performance parameters of the insulating adhesive to be tested to obtain an initial data set, wherein the initial data set includes a resistivity sequence, a dielectric constant sequence, and a tensile strength sequence; Extracting the characteristic vector of the initial data set, and constructing a matrix based on the characteristic vector to obtain a performance characteristic matrix; Based on the performance characteristic matrix, a dynamic loading experiment is performed on the insulating adhesive to be tested to generate a dynamic response data set; Constructing a time series curve of the dynamic response data set, performing degradation trend fitting processing on the time series curve, and obtaining a degradation parameter set; Performing a multi-dimensional evaluation on the degradation parameter set based on information entropy to obtain a comprehensive score; Anomaly detection is performed on the comprehensive score, contribution analysis is performed on the comprehensive score after detection and the degradation parameter set, and the analysis result is iteratively divided into quality levels to obtain a quality judgment result.

[0007] Furthermore, the step of synchronously collecting multi-dimensional performance parameters of the insulating adhesive to be tested to obtain an initial data set, wherein the initial data set includes a resistivity sequence, a dielectric constant sequence, and a tensile strength sequence, comprises: Perform multi-channel physical excitation on the insulating adhesive to be tested, and perform time stamp alignment processing on the excitation signal to obtain the original measurement signal; Performing signal segmentation on the original measurement signal, and performing multi-dimensional parameter conversion processing on the segmented signal to obtain preliminary performance data, wherein the preliminary performance data includes resistivity, dielectric constant and tensile strength; Performing dynamic filtering processing on the preliminary performance data to obtain a smoothed data sequence, and performing statistical analysis on the smoothed data sequence based on a sliding window to obtain a screened data set; Segmenting the screened data set into time series, and performing feature labeling processing on the segmentation results to obtain a segmented feature data set; The segmented feature data sets are integrated in the order of timestamps to obtain an initial data set.

[0008] Furthermore, the step of extracting the feature vector of the initial data set and constructing a matrix based on the feature vector to obtain a performance feature matrix includes: Performing multi-scale hierarchical processing on the initial data set based on the resistivity sequence, the dielectric constant sequence and the tensile strength sequence to obtain a hierarchical sequence group; Performing nonlinear mapping on the hierarchical sequence group to obtain a primary feature vector set, and performing weighted fusion processing on the hierarchical sequence group based on the primary feature vector set to obtain a fusion feature matrix; Performing singular value decomposition processing on the fused feature matrix to obtain a reduced-dimensional feature matrix; Based on the projection value of each dimension in the dimension-reduced feature matrix, the primary feature vector set is subjected to residual correction processing to obtain an optimized feature vector set; The optimized feature vector set is subjected to tensor reconstruction processing to obtain a performance feature matrix.

[0009] Furthermore, the step of performing a dynamic loading experiment on the insulating adhesive to be tested based on the performance characteristic matrix to generate a dynamic response data set includes: Performing loading parameter mapping processing on the performance characteristic matrix to obtain a loading condition sequence; Performing a progressive multi-field loading experiment on the insulating adhesive to be tested according to the loading condition sequence to obtain an initial response data set; Performing multi-scale time-frequency decomposition processing on the initial response data set to obtain a time-frequency feature component set; Performing adaptive noise reduction processing on the initial response data set according to the time-frequency feature component set to obtain a smoothed response data set; Performing state transition probability modeling on the smooth response data set, constructing a state transition relationship between parameters in the smooth response data set, and obtaining a dynamic evolution feature matrix; The smooth response data set is subjected to multi-parameter fusion processing according to the dynamic evolution characteristic matrix to obtain a dynamic response data set.

[0010] Furthermore, the step of constructing a time series curve of the dynamic response data set, performing degradation trend fitting processing on the time series curve, and obtaining a degradation parameter set includes: Performing multi-dimensional time series decomposition processing on the dynamic response data set to obtain a primary time series group; Performing curve smoothing processing on the primary time series group, and marking the smoothing results in sections to obtain a section trend series set; Setting the initialization decay rate of each trend segment in the segmented trend sequence set, iteratively optimizing each of the initialization decay rates based on an optimization algorithm, and obtaining a fitting parameter initial value set; Performing dynamic weight fitting processing on the segmented trend sequence set according to the initial value set of the fitting parameters to obtain an optimized time series curve; Extracting degradation features of the optimized time series curve, and reorganizing the degradation features according to performance dimensions to obtain a degradation feature matrix; A multi-scale residual analysis is performed on the degradation feature matrix, and the degradation feature matrix is ​​optimized based on the distribution characteristics of the residual results to obtain a degradation parameter set.

[0011] Furthermore, the step of performing multi-dimensional evaluation on the degradation parameter set based on information entropy to obtain a comprehensive score includes: Performing single feature extraction on the degradation parameter set to obtain a single performance indicator set, and performing probability distribution calculation on information entropy based on the single performance indicator set to obtain an entropy value sequence; Setting an initial weight vector, and performing nonlinear mapping processing on the initial weight vector according to each entropy value in the entropy value sequence to obtain an optimized weight set; Performing weighted fusion processing on the individual performance indicator sets according to the optimization weight set to obtain a preliminary evaluation score; Performing residual analysis on the preliminary evaluation score to obtain an optimized evaluation score; A comprehensive characterization function is constructed based on the optimization evaluation score, and the optimization evaluation score is subjected to comprehensive characterization processing according to the comprehensive characterization function to obtain a comprehensive score.

[0012] Furthermore, the step of performing abnormality detection on the comprehensive score, performing contribution analysis on the detected comprehensive score and the degradation parameter set, and iteratively dividing the analysis result into quality grades to obtain a quality determination result includes: Performing multi-scale sliding window processing on the comprehensive score to obtain a preliminary abnormal distribution sequence; According to the statistical distribution characteristics of the abnormal preliminary distribution sequence, dynamic anomaly detection is performed on the comprehensive score to obtain an abnormal label set; According to the distribution density and abnormal degree of the abnormal points in the abnormal mark set, weighted cluster analysis is performed on the abnormal points in the abnormal mark set to obtain an abnormal cluster feature vector; Decomposing the degradation parameter set according to the abnormal clustering feature vector to obtain a parameter influence weight matrix; Constructing a multidimensional projection transformation matrix based on the parameter influence weight matrix, mapping the comprehensive score and the degradation parameter set to a high-dimensional feature space through the multidimensional projection transformation matrix to obtain a quality feature projection set; The quality feature projection set is iteratively optimized based on a support vector machine classification algorithm to obtain a quality determination result, which includes an abnormal mark and a multi-level quality classification label.

[0013] The present invention also provides a testing device for insulating glue, comprising: An acquisition module is used to synchronously acquire multi-dimensional performance parameters of the insulating adhesive to be tested to obtain an initial data set, wherein the initial data set includes a resistivity sequence, a dielectric constant sequence, and a tensile strength sequence; An extraction module, used to extract the feature vector of the initial data set, and construct a matrix based on the feature vector to obtain a performance feature matrix; An experimental module, used for performing a dynamic loading experiment on the insulating adhesive to be tested based on the performance characteristic matrix to generate a dynamic response data set; A construction module is used to construct a time series curve of the dynamic response data set, and perform degradation trend fitting processing on the time series curve to obtain a degradation parameter set; An evaluation module, used for performing a multi-dimensional evaluation on the degradation parameter set based on information entropy to obtain a comprehensive score; The determination module is used to perform anomaly detection on the comprehensive score, perform contribution analysis on the detected comprehensive score and the degradation parameter set, and iteratively divide the analysis result into quality levels to obtain a quality determination result.

[0014] The present invention further provides a computer device, which includes a processor, a memory, and a computer program stored in the memory and running on the processor, and the processor implements the above method when executing the computer program.

[0015] The present invention also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the above method is implemented.

[0016] Beneficial effects: In the present invention, by synchronously collecting the multi-dimensional performance parameters of the insulating glue to be tested, a performance characteristic matrix is ​​constructed, so that the test data can fully reflect the performance of the insulating glue in different dimensions. Based on the performance characteristic matrix, a dynamic loading experiment is implemented to generate a dynamic response data set, so that the test method not only focuses on the initial performance indicators of the insulating glue, but also can capture the response characteristics of the material under dynamic use conditions, thereby more accurately reflecting its long-term stability. A time series curve of the dynamic response data set is further constructed, and the curve is subjected to degradation trend fitting processing to form a degradation parameter set, so as to achieve quantitative analysis of the performance change trend of the insulating glue, which is helpful to accurately predict its durability. In addition, information entropy is used to perform multi-dimensional evaluation of the degradation parameter set to generate a comprehensive score, so that the quality judgment evaluates the overall quality level of the insulating glue through the fusion of multi-dimensional information, thereby improving the scientificity and accuracy of the test. At the same time, based on anomaly detection, the comprehensive score is analyzed, and combined with the contribution of the degradation parameter set, potential quality risks can be effectively identified to avoid misjudgment caused by abnormal local indicators. Further, by iteratively dividing the quality level, it is ensured that the quality evaluation results can more reasonably guide the selection and application of the insulating glue, and improve the reliability and service life of the insulating glue in actual engineering. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 It is a schematic diagram of the steps of a testing method of an insulating adhesive of the present invention; Figure 2 It is a structural schematic block diagram of a testing device for insulating glue of the present invention; Figure 3 is a schematic block diagram of the structure of a computer device according to an embodiment of the present invention; The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0018] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.

[0019] In the description of the present invention, it should be understood that the terms "center", "longitudinal", "lateral", "length", "width", "thickness", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", "clockwise", "counterclockwise", etc., indicating the orientation or positional relationship, are based on the orientation or positional relationship shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first" and "second" are only used for descriptive purposes, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Therefore, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. In the description of the present invention, the meaning of "multiple" is two or more, unless otherwise clearly and specifically defined.

[0020] In the description of the present invention, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected" and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection, a direct connection, or an indirect connection through an intermediate medium, it can be the internal connection of two elements or the interaction relationship between two elements. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0021] In the present invention, unless otherwise clearly specified and limited, a first feature being "above" or "below" a second feature may include that the first and second features are in direct contact, or may include that the first and second features are not in direct contact but are in contact through another feature between them. Moreover, a first feature being "above", "above" and "above" a second feature includes that the first feature is directly above and obliquely above the second feature, or simply indicates that the first feature is higher in level than the second feature. A first feature being "below", "below" and "below" a second feature includes that the first feature is directly below and obliquely below the second feature, or simply indicates that the first feature is lower in level than the second feature.

[0022] Reference Figure 1 The present invention provides a method for testing insulating glue, comprising the following steps: S1: synchronously collecting multi-dimensional performance parameters of the insulating adhesive to be tested to obtain an initial data set, wherein the initial data set includes a resistivity sequence, a dielectric constant sequence, and a tensile strength sequence; In step S1, the sample preparation of the insulating glue to be tested is clarified, and the insulating glue is prepared into a standardized specimen, such as a rectangular sheet or a cylindrical test block, to ensure the consistency of the test conditions. Use high-precision testing equipment to synchronously collect multi-dimensional performance parameters, that is, complete the measurement of multiple parameters in the same time period to avoid data deviations caused by time differences. Specifically, a precision resistivity tester is used to measure the resistivity sequence. The resistivity is calculated by applying a constant voltage and recording the current. It is necessary to repeat the measurement multiple times under different temperature or humidity conditions to form a resistivity sequence containing multiple data points. At the same time, the acquisition of the dielectric constant sequence is tested by a dielectric spectrometer. The change trend of the dielectric constant is obtained by measuring the capacitance characteristics of the insulating glue at different frequencies. Assuming that the test frequency range is set to 1Hz to 1MHz, a dielectric constant sequence that changes with frequency can be obtained. The measurement of the tensile strength sequence depends on the universal material testing machine. By applying a gradually increasing tensile force to the sample until it breaks, the maximum strength value of each tensile test is recorded, and a tensile strength sequence is formed after repeated multiple times. Generate multidimensional data containing multiple groups of time series, and then perform sliding window anomaly removal on the multidimensional data. The window size is set to 5 consecutive sampling points. The outliers are judged and removed based on the statistical standard deviation threshold to obtain the initial data set, which includes the screened resistivity series, dielectric constant series and tensile strength series. For example, when testing a piece of silicone-based insulating glue, the resistivity series may show 10^12 Ω·m at 25℃ and drop to 10^11Ω·m at 50℃, while the dielectric constant series is 3.5 at 1kHz and 3.2 at 100kHz, and the tensile strength series may contain the results of 5 measurements, such as 20MPa, 22MPa, etc. The test data is integrated into the initial data set. The entire process requires strict control of environmental variables and instrument calibration to ensure the repeatability and scientificity of the data.

[0023] S2: extracting the feature vector of the initial data set, and constructing a matrix based on the feature vector to obtain a performance feature matrix; In step S2, the initial data set is normalized to eliminate dimensional differences, the mean, variance and skewness of each group of sequences are calculated, and a eigenvector containing statistical features is generated. The normalized sequence and the eigenvector are then combined into a multidimensional matrix, and then the multidimensional matrix is ​​subjected to principal component analysis and dimensionality reduction processing, and the principal components whose cumulative contribution rate reaches a preset ratio are retained to obtain a performance characteristic matrix, which characterizes the multidimensional performance distribution characteristics of the insulating glue. Specifically, the initial data set needs to be preprocessed, including removing outliers or noise, extracting eigenvectors for each parameter sequence, calculating the mean, variance and maximum value of the resistivity sequence, and obtaining a three-dimensional eigenvector such as [10^12, 10^2, 10^13]. Similarly, the dielectric constant sequence may extract the frequency response slope and average value, and the tensile strength sequence extracts the peak value and coefficient of variation. After extracting the eigenvectors, these vectors are combined into a matrix. For example, if there are 10 samples, each sample extracts the eigenvectors of 3 parameters, and each vector has 3 components, then a 10×9 performance characteristic matrix can be constructed, where each row represents the characteristics of a sample, and each column corresponds to a certain eigenvalue of a certain parameter. Assuming that 5 silicone samples are tested, the resistivity eigenvector may be [mean 10^12, variance 10^3, maximum value 10^13], the dielectric constant eigenvector is [mean 3.5, slope 0.01, minimum value 3.0], and the tensile strength eigenvector is [mean 21, variance 2, peak value 23]. Arrange these vectors in rows and expand them into columns to form a 5×9 matrix. After the matrix is ​​constructed, principal component analysis (PCA) is used to further reduce the dimension or verify the significance of the data. The whole process is implemented with the help of programming tools such as MATLAB or Python to improve computational efficiency and accuracy.

[0024] S3: performing a dynamic loading experiment on the insulating adhesive to be tested based on the performance characteristic matrix to generate a dynamic response data set; In step S3, the performance characteristic matrix is ​​subjected to loading parameter mapping processing to determine the range of electric field and stress increment, and a progressive electric field and mechanical stress loading experiment is performed on the insulating glue sample. The multi-dimensional response data of resistivity change, dielectric constant offset and deformation are recorded in real time, and then the response data is smoothed by Kalman filtering. The noise effect is eliminated by iteratively updating the state estimation to obtain a dynamic data set, which contains a multi-parameter change sequence in each loading stage. Specifically, the implementation process requires the design of a dynamic loading experiment scheme, and the servo control system is used to apply periodic stress or electric field loading to the sample. The loading conditions can be sinusoidal waveforms or step increments, depending on the application scenario of the insulating glue. The performance characteristic matrix is ​​used as an input parameter to guide the selection of experimental conditions. If the matrix shows that the tensile strength characteristics of a sample are low, the loading frequency is reduced to avoid premature failure. During the experiment, the response data of the sample is recorded in real time, such as the change of resistivity over time, the drift of dielectric constant with frequency, or the attenuation of tensile strength with the number of cycles. These data constitute a dynamic response data set. Assuming that 1000 tensile cycles are applied to the silicone sample, each with a tensile force of 15MPa, the resistivity is recorded to gradually decrease from 10^12 Ω·m to 10^10 Ω·m, and the dielectric constant changes from 3.5 to 3.8, indicating that microstructural damage may have occurred inside the material. Dynamic loading experiments require the use of high-precision sensors and data acquisition systems to ensure the continuity and resolution of the response data. At the same time, the experimental design should consider the effects of variables such as loading rate and ambient temperature. Finally, the dynamic response data set is stored in the form of a multidimensional time series.

[0025] S4: constructing a time series curve of the dynamic response data set, performing degradation trend fitting processing on the time series curve, and obtaining a degradation parameter set; In step S4, the dynamic data set is processed for time series curve generation, and the change curves of resistivity, dielectric constant and tensile strength are constructed respectively. Then, an exponential decay model is fitted for each curve, and the fitting parameters are iteratively calculated by the least square method. Then, an adaptive weight factor is introduced according to the variance of each parameter change to adjust the fitting process, and a degradation parameter set is obtained. The degradation parameter set includes the decay rate and stable value parameters of each performance index, which reveals the stability of the material in long-term use. First, the dynamic response data set is sorted into curves in chronological order. The change curve of resistivity with loading cycle can be represented by a line graph. Assuming that the resistivity decreases from 10^12 Ω·m to 10^10 Ω·m after 1000 cycles, the curve shows an exponential decay trend. Fitting processing is performed on these time series curves, including linear regression, exponential decay or polynomial fitting. The specific choice depends on the shape of the curve. For example, the exponential model may be more suitable for resistivity decay, while the dielectric constant requires quadratic polynomial fitting. Through fitting, the degradation parameter set is extracted, including the decay rate, steady-state value or inflection point time. Assuming that the resistivity fitting result is R(t)=10^12·e^(-0.002t), the decay rate of 0.002 becomes one of the degradation parameters. The fitting process needs to be completed with the help of mathematical software Origin or R language, and the goodness of fit is verified by residual analysis.

[0026] S5: performing a multi-dimensional evaluation on the degradation parameter set based on information entropy to obtain a comprehensive score; In step S5, the degradation parameter set is processed by scoring function calculation, and the single score of each performance index is generated based on the decay rate and the stable value. A multidimensional evaluation function including the initial weight is designed to combine the single score, and then the degradation parameter set is processed by information entropy analysis. The entropy value is calculated by probability distribution and the weight of each index is dynamically adjusted. The adjusted weight is then applied to the evaluation function to obtain a comprehensive score, which characterizes the overall performance level of the insulating glue. Specifically, the information entropy of the degradation parameter set is calculated. The information entropy reflects the degree of discreteness of the parameter distribution. For example, if the resistivity decay rate fluctuates less between [0.002, 0.0018, 0.0022], the entropy value is low, indicating that the degradation consistency is high. The entropy value of each parameter is weighted and combined. The weight can be determined according to the application requirements. The resistivity can be set to 40%, the dielectric constant and the tensile strength can each account for 30%, and the comprehensive score can be obtained by weighted summation. Assuming that the entropy values ​​of the five samples are [0.1, 0.15, 0.12, 0.09, 0.11], the weighted comprehensive score may be [0.85, 0.78, 0.82, 0.88, 0.84]. If the resistivity degradation entropy of a silicone sample is low and the dielectric constant change entropy is high, the comprehensive score may be low, reflecting its uneven performance. The entire evaluation process requires clarifying the entropy calculation formula and ensuring data standardization. The entire step achieves unified quantification of multi-dimensional parameters through information entropy.

[0027] S6: Perform anomaly detection on the comprehensive score, perform contribution analysis on the detected comprehensive score and the degradation parameter set, and iteratively divide the analysis result into quality levels to obtain a quality determination result.

[0028] In step S6, a threshold comparison process is performed on the comprehensive score, a quality threshold is set based on historical data statistics and anomalies are marked, a contribution analysis is performed on the degradation parameter set and the comprehensive score, the relative influence ratio of each parameter on the score is calculated, and then the degradation parameter set and the comprehensive score are input into the support vector machine classification model. The quality level is divided iteratively through hyperplane optimization to obtain a quality judgment result, which includes anomaly marking and quality level classification.

[0029] Specifically, the comprehensive score is first detected for anomalies, and the box plot method is used to identify outliers in the score. If 0.5 out of the 5 scores is significantly lower, it is marked as an anomaly. Contribution analysis is performed to calculate the contribution of each degradation parameter to the comprehensive score. For example, the partial regression coefficient is used to find that resistivity attenuation has the greatest impact on the score. Finally, the quality level is iteratively divided. The score threshold [0.9, 0.8, 0.7] can be set to correspond to excellent, good, and medium levels. The threshold is repeatedly adjusted and the classification effect is verified to finally determine the quality of the sample.

[0030] In one embodiment, the step of synchronously collecting multi-dimensional performance parameters of the insulating adhesive to be tested to obtain an initial data set, wherein the initial data set includes a resistivity sequence, a dielectric constant sequence, and a tensile strength sequence, comprises: Perform multi-channel physical excitation on the insulating adhesive to be tested, and perform time stamp alignment processing on the excitation signal to obtain the original measurement signal; Performing signal segmentation on the original measurement signal, and performing multi-dimensional parameter conversion processing on the segmented signal to obtain preliminary performance data, wherein the preliminary performance data includes resistivity, dielectric constant and tensile strength; Performing dynamic filtering processing on the preliminary performance data to obtain a smoothed data sequence, and performing statistical analysis on the smoothed data sequence based on a sliding window to obtain a screened data set; Segmenting the screened data set into time series, and performing feature labeling processing on the segmentation results to obtain a segmented feature data set; The segmented feature data sets are integrated in the order of timestamps to obtain an initial data set.

[0031] In the above embodiment, the insulating rubber sample to be tested is subjected to parallel processing of electric field excitation, dielectric response measurement and mechanical tensile loading, and the current response is recorded by applying a preset gradient voltage to generate an electric signal sequence, the dielectric response is recorded by high-frequency capacitance scanning to generate a capacitance signal sequence, and the deformation is recorded by gradually increasing the tensile stress to generate a stress signal sequence, and then the above signals are time-stamp aligned to synchronize multi-channel data to obtain the original measurement signal, which includes unprocessed current, capacitance and stress time series. The original measurement signal is subjected to signal segmentation and feature calculation processing, the current sequence is converted into a resistivity sequence by Ohm's law calculation, the capacitance sequence is converted into a dielectric constant sequence by a capacitance-dielectric model, and the stress sequence is converted into a tensile strength sequence by a stress-strain relationship, and then the converted sequence is subjected to sampling rate standardization processing to unify the time resolution, and preliminary performance data of a preliminary numerical sequence including resistivity, dielectric constant and tensile strength is obtained.

[0032] The preliminary performance data are subjected to wavelet transform decomposition processing, and each sequence is decomposed into high-frequency and low-frequency components. Then, the high-frequency components are subjected to adaptive threshold filtering to suppress noise, and the low-frequency components are reconstructed and merged to generate a smooth curve. Then, the filtering parameters are dynamically adjusted according to the sampling point density of the preliminary performance data to obtain a smooth data sequence, which includes the resistivity, dielectric constant and tensile strength sequences after denoising. The smooth data sequence is subjected to sliding window statistical analysis processing, and the window length is set to 7 consecutive sampling points. The mean and standard deviation in each window are calculated, and then the abnormal points are marked and eliminated according to the three-times standard deviation criterion. Then, the eliminated sequences are interpolated and filled to maintain data continuity, and a screening data set is obtained, which includes the resistivity, dielectric constant and tensile strength sequences after abnormal elimination.

[0033] The screening data set is processed by change point detection. The performance mutation point is determined by calculating the slope change rate of adjacent data points and the sequence is segmented. Then, the statistical features of each segmented sequence are calculated, including the mean, peak value and change trend slope. Then, time interval labels are added to each sequence according to the segmented features to obtain a segmented feature data set, which includes annotated resistivity, dielectric constant and tensile strength segmented sequences. The segmented feature data set is processed by multidimensional vector reorganization, and the segmented sequences of resistivity, dielectric constant and tensile strength are merged into a unified matrix in timestamp order. Then, the matrix is ​​column normalized to balance the dimensional differences of each parameter. Then, metadata labels are added to the matrix according to the experimental conditions, including sampling frequency and excitation parameters, to obtain the initial data set.

[0034] In one example, the step of extracting the feature vector of the initial data set and constructing a matrix based on the feature vector to obtain a performance feature matrix includes: Performing multi-scale hierarchical processing on the initial data set based on the resistivity sequence, the dielectric constant sequence and the tensile strength sequence to obtain a hierarchical sequence group; Performing nonlinear mapping on the hierarchical sequence group to obtain a primary feature vector set, and performing weighted fusion processing on the hierarchical sequence group based on the primary feature vector set to obtain a fusion feature matrix; Performing singular value decomposition processing on the fused feature matrix to obtain a reduced-dimensional feature matrix; Based on the projection value of each dimension in the dimension-reduced feature matrix, the primary feature vector set is subjected to residual correction processing to obtain an optimized feature vector set; The optimized feature vector set is subjected to tensor reconstruction processing to obtain a performance feature matrix.

[0035] In the above embodiment, according to the resistivity sequence, dielectric constant sequence and tensile strength sequence in the initial data set, each sequence is decomposed by wavelet transform, high-frequency components and low-frequency components are extracted, and corresponding multi-scale subsequences are generated respectively, and then all subsequences are rearranged and combined according to the frequency level to obtain a hierarchical sequence group, which characterizes the distribution characteristics of the performance parameters of the insulating glue at different time scales. Kernel function mapping is performed on each subsequence in the hierarchical sequence group, and the Gaussian kernel function is used to project the subsequence into a high-dimensional feature space, and then the Euclidean distance and cosine similarity of each group of subsequences in the high-dimensional space are calculated, and the distance and similarity are combined into a primary feature vector to obtain a primary feature vector set, which characterizes the nonlinear correlation characteristics of the performance of the insulating glue.

[0036] According to the Euclidean distance value of each feature vector in the primary feature vector set, a dynamic weight coefficient is assigned to the corresponding subsequence in the hierarchical sequence group, and all subsequences are fused and calculated using the weighted average method to generate a fused sequence containing multidimensional performance parameters. The fused sequence is then reconstructed into a two-dimensional matrix according to the time dimension and the parameter dimension to obtain a fused feature matrix, which characterizes the comprehensive distribution characteristics of the insulating adhesive performance. The fused feature matrix is ​​subjected to singular value decomposition to extract the first k singular values ​​and their corresponding left and right singular vectors, where k is determined by the cumulative singular value contribution rate reaching a preset threshold. Subsequently, the fused feature matrix is ​​linearly projected according to the extracted singular vectors to generate a feature representation after dimensionality reduction, and a reduced dimensionality feature matrix is ​​obtained, which characterizes the main change patterns of the insulating adhesive performance.

[0037] According to the projection value of each dimension in the dimension reduction feature matrix, the residual calculation is performed on the corresponding feature vector in the primary feature vector set, and the least square method is used to fit the mapping relationship between the residual and the projection value. Then, the primary feature vector is iteratively updated according to the mapping relationship to obtain an optimized feature vector set, which characterizes the multidimensional statistical characteristics of the performance of the insulating glue. The outer product operation is performed on each feature vector in the optimized feature vector set to generate the corresponding feature tensor sub-block, and then all feature tensor sub-blocks are stacked into a three-dimensional tensor according to the time dimension, parameter dimension and statistical dimension, and then the three-dimensional tensor is standardized to obtain a performance feature matrix, which characterizes the global distribution characteristics of the multi-dimensional performance of the insulating glue.

[0038] In one example, the step of performing a dynamic loading experiment on the insulating adhesive to be tested based on the performance characteristic matrix to generate a dynamic response data set includes: Performing loading parameter mapping processing on the performance characteristic matrix to obtain a loading condition sequence; Performing a progressive multi-field loading experiment on the insulating adhesive to be tested according to the loading condition sequence to obtain an initial response data set; Performing multi-scale time-frequency decomposition processing on the initial response data set to obtain a time-frequency feature component set; Performing adaptive noise reduction processing on the initial response data set according to the time-frequency feature component set to obtain a smoothed response data set; Performing state transition probability modeling on the smooth response data set, constructing a state transition relationship between parameters in the smooth response data set, and obtaining a dynamic evolution feature matrix; The smooth response data set is subjected to multi-parameter fusion processing according to the dynamic evolution characteristic matrix to obtain a dynamic response data set.

[0039] In the above embodiment, a condition sequence suitable for dynamic loading experiments is generated through a performance characteristic matrix (including multidimensional characteristic vectors such as resistivity, dielectric constant, tensile strength, etc.) to ensure that the loading range matches the material performance. Specifically, the statistical distribution of each characteristic vector is extracted from the performance characteristic matrix, the increasing range of electric field intensity (such as 0 to 10 kV / mm) is determined according to the resistivity sequence, the increasing range of mechanical stress (such as 0 to 5 MPa) is determined according to the tensile strength sequence, and the loading time step (such as 0.1 second) is determined according to the dielectric constant sequence. The above ranges are discretized into increasing steps to generate a loading condition sequence.

[0040] According to the sequence of loading conditions, a progressive multi-field loading experiment is carried out on the insulating glue to be tested. At this stage, the experiment needs to be completed under the coupling of multiple fields (such as thermal, mechanical, electric fields, etc.) to simulate the complex working conditions of the insulating glue in actual application scenarios. In the experimental setting, it is necessary to accurately control the step-by-step changes of the loading conditions so that the test can be gradually promoted from low to high and from simple to complex, ensuring that the experimental results can capture the response characteristics under low load conditions and reflect the failure behavior under high load or extreme working conditions. During the experiment, the original response data of the insulating glue under the loading conditions are obtained through high-precision sensors and sampling devices, including multi-dimensional information such as time, displacement, temperature, and current, and finally form an initial response data set.

[0041] The initial response data set is subjected to multi-scale time-frequency decomposition to extract the time-frequency characteristic component set, and the multi-scale laws in the data are revealed through signal decomposition technology. Discrete wavelet transform (DWT) is applied to the resistivity, dielectric constant and deformation series respectively, and the Daubechies wavelet basis is selected and decomposed into 5 layers to obtain low-frequency trend components (such as the long-term downward trend of resistivity) and high-frequency detail components (such as instantaneous fluctuations in deformation). The amplitude and phase of each component are extracted and integrated into the time-frequency characteristic component set.

[0042] Based on the time-frequency characteristic component set, the initial response data set is adaptively denoised to generate a smooth response data set. The noise threshold is determined by the amplitude of the high-frequency component (for example, an amplitude less than 0.1 is considered as noise), and then the Kalman filter is applied to iteratively update the data. The high-frequency noise in the resistivity series can be suppressed, and its main downward trend can be retained. Finally, a smooth response data set is output, and the random interference in the experiment is removed. Subsequently, the state transition probability modeling of the smooth data set is performed, and the conditional probability between resistivity, dielectric constant and deformation is calculated by the hidden Markov model (HMM) (for example, the probability of dielectric constant increasing when resistivity decreases is 0.8), and a dynamic evolution characteristic matrix is ​​constructed to reflect the dynamic dependency between parameters.

[0043] The smooth response data set is processed by multi-parameter fusion according to the dynamic evolution characteristic matrix to generate the final dynamic response data set. The transition probability of resistivity and dielectric constant is used as the weight, and the smooth data is weighted and normalized, and its statistical distribution (such as mean and standard deviation) is calculated and integrated into a multidimensional dynamic response data set, which not only contains the change sequence of each parameter, but also reflects its evolution law, providing a comprehensive basis for the performance evaluation of insulating adhesives.

[0044] In one example, the step of constructing a time series curve of the dynamic response data set, performing degradation trend fitting processing on the time series curve, and obtaining a degradation parameter set includes: Performing multi-dimensional time series decomposition processing on the dynamic response data set to obtain a primary time series group; Performing curve smoothing processing on the primary time series group, and marking the smoothing results in sections to obtain a section trend series set; Setting the initialization decay rate of each trend segment in the segmented trend sequence set, iteratively optimizing each of the initialization decay rates based on an optimization algorithm, and obtaining a fitting parameter initial value set; Performing dynamic weight fitting processing on the segmented trend sequence set according to the initial value set of the fitting parameters to obtain an optimized time series curve; Extracting degradation features of the optimized time series curve, and reorganizing the degradation features according to performance dimensions to obtain a degradation feature matrix; A multi-scale residual analysis is performed on the degradation feature matrix, and the degradation feature matrix is ​​optimized based on the distribution characteristics of the residual results to obtain a degradation parameter set.

[0045] In the above embodiment, the dynamic response data set is subjected to signal hierarchical decomposition processing, and the response data of resistivity, dielectric constant and tensile strength are respectively extracted as independent time series, and then each series is segmented in the frequency domain by short-time Fourier transform to separate the low-frequency trend component and the high-frequency fluctuation component, and then the low-frequency trend component is reconstructed into a continuous curve along the time axis to obtain a primary time series group, which includes the low-frequency trend series of resistivity, dielectric constant and tensile strength.

[0046] The primary time series group is subjected to moving average filtering, and the sliding window length is set to 10 sampling points to smooth local fluctuations. Then, the trend turning points are detected according to the rate of change of the second-order derivative of each sequence. Each sequence is divided into multiple sub-intervals, and then the start time and end time labels are added to each sub-interval to obtain a segmented trend sequence set, which includes resistivity, dielectric constant and tensile strength trend segments with time annotations.

[0047] The exponential decay model is pre-fitted for the segmented trend sequence set, and the decay rate and stable value parameters are initialized for each trend segment. Then, the optimal initial value of each segment is calculated by the Newton iteration method, and the parameter range is adjusted by introducing constraints according to the slope continuity between segments to obtain the initial value set of fitting parameters, which includes the preliminary degradation parameters of each segment of resistivity, dielectric constant and tensile strength. The segmented trend sequence set is fitted by the weighted least squares method according to the initial value set of fitting parameters, and the residual variance is calculated for each segment and an adaptive weight factor is assigned according to the size of the variance. Then, the fitting parameters are iteratively updated until the residual converges, and then the fitting curves of each segment are spliced ​​in chronological order and the transition at the connection is smoothed to obtain the optimized time series curve, which includes the resistivity, dielectric constant and tensile strength degradation curves after continuous fitting.

[0048] The optimized time series curves are subjected to characteristic decomposition processing, and the attenuation rate, stable value and inflection point time of each curve are calculated. Then, multidimensional statistical analysis is performed on each parameter, and the skewness and kurtosis indicators are introduced to characterize the degradation distribution characteristics. All characteristic parameters are then reorganized into a matrix form according to the performance dimension to obtain a degradation characteristic parameter matrix, which contains the degradation characteristic vectors of resistivity, dielectric constant and tensile strength.

[0049] The degradation characteristic parameter matrix is ​​subjected to multi-scale residual analysis, and the characteristic parameters in the matrix are substituted into the long-period and short-period degradation models for verification. Subsequently, the parameter weights are adjusted according to the distribution characteristics of the verification residuals, and the confidence intervals of each parameter are iteratively optimized through Monte Carlo simulation. Finally, all refined parameters are integrated to obtain the degradation parameter set, which includes the final attenuation rate and stable value parameters of resistivity, dielectric constant and tensile strength.

[0050] In another embodiment, the calculation expression of the above content is: ; in, is the degradation parameter set; To accumulate the results of all K trend segments; is the dynamic weight factor, the adaptive weight of the kth trend segment; is an exponential decay term, indicating the degradation trend of the kth trend segment, is the decay rate of the segment, is a time variable; is the inverse Fourier transform term, represents the frequency domain signal of the kth segment, represents the inverse Fourier transform, which reconstructs the low-frequency trend component into a time series; It is the Hadamard product, element-by-element multiplication, which means the element-by-element multiplication of two matrices or vectors; is the second-order derivative term, is the moving average filtering result with a sliding window length of 10, and the second-order derivative Calculate its rate of change; F represents the residual analysis and distribution characteristic optimization function, specifically: , is the residual mean calculated by multi-scale residual analysis, reflecting the deviation of long-period and short-period model verification; is the residual adjustment coefficient, used to balance the residual variance Contributions to the formula,optimized through Monte Carlo simulation; is the residual variance of each segment in the fitting process, which measures the error distribution of the fitting curve; is the statistical distribution of the degenerate characteristic matrix, is the skewness of the degradation distribution, which characterizes the asymmetry of the degradation characteristics.

[0051] In one example, the step of performing a multi-dimensional evaluation on the degradation parameter set based on information entropy to obtain a comprehensive score includes: Performing single feature extraction on the degradation parameter set to obtain a single performance indicator set, and performing probability distribution calculation on information entropy based on the single performance indicator set to obtain an entropy value sequence; Setting an initial weight vector, and performing nonlinear mapping processing on the initial weight vector according to each entropy value in the entropy value sequence to obtain an optimized weight set; Performing weighted fusion processing on the individual performance indicator sets according to the optimization weight set to obtain a preliminary evaluation score; Performing residual analysis on the preliminary evaluation score to obtain an optimized evaluation score; A comprehensive characterization function is constructed based on the optimization evaluation score, and the optimization evaluation score is subjected to comprehensive characterization processing according to the comprehensive characterization function to obtain a comprehensive score.

[0052] In the above embodiment, according to the time series data in the degradation parameter set, the time series curve of each degradation parameter is subjected to feature decomposition processing, and the feature parameters of the attenuation rate, the stable value and the fluctuation amplitude are extracted to obtain a single performance index set, wherein the single performance index set includes the feature values ​​of the resistivity degradation index, the dielectric constant degradation index and the tensile strength degradation index. Each feature value in the single performance index set is normalized, and the probability distribution function of each feature value is calculated. According to the probability distribution function, the information entropy is discretized and calculated to obtain an entropy value sequence, wherein the entropy value sequence characterizes the information uncertainty distribution of each degradation parameter.

[0053] According to each entropy value in the entropy value sequence, a nonlinear mapping process is performed on the corresponding weight in the initial weight vector, and an adjustment coefficient is calculated using the inverse relationship between the entropy value and the weight. The initial weight vector is iteratively updated by the adjustment coefficient to obtain an optimized weight set, wherein the optimized weight set reflects the contribution of each degradation parameter to the comprehensive evaluation. According to the weight value in the optimized weight set, a weighted summation process is performed on the eigenvalues ​​in the single performance indicator set, a multidimensional evaluation function is constructed, and a linear combination calculation is performed on the single performance indicator set by the multidimensional evaluation function to obtain a preliminary evaluation score, wherein the preliminary evaluation score represents the preliminary quantitative result of the performance of the insulating adhesive.

[0054] The residual calculation processing is performed on the preliminary evaluation score and the expected value of the degradation parameter set, and a residual distribution model is constructed. The preliminary evaluation score is iteratively corrected according to the residual distribution model, and the corrected score sequence is smoothed and filtered to obtain an optimized evaluation score, wherein the optimized evaluation score has higher stability and consistency. According to the optimized evaluation score, the multi-dimensional degradation characteristics of the degradation parameter set are normalized and mapped, and a comprehensive characterization function is constructed. The optimized evaluation score is nonlinearly transformed and integrally calculated through the comprehensive characterization function to obtain a comprehensive score, wherein the comprehensive score characterizes the overall performance level of the insulating adhesive.

[0055] In one embodiment, the step of performing abnormality detection on the comprehensive score, performing contribution analysis on the detected comprehensive score and the degradation parameter set, and iteratively dividing the analysis result into quality grades to obtain a quality determination result includes: Performing multi-scale sliding window processing on the comprehensive score to obtain a preliminary abnormal distribution sequence; According to the statistical distribution characteristics of the abnormal preliminary distribution sequence, dynamic anomaly detection is performed on the comprehensive score to obtain an abnormal label set; According to the distribution density and abnormal degree of the abnormal points in the abnormal mark set, weighted cluster analysis is performed on the abnormal points in the abnormal mark set to obtain an abnormal cluster feature vector; Decomposing the degradation parameter set according to the abnormal clustering feature vector to obtain a parameter influence weight matrix; Constructing a multidimensional projection transformation matrix based on the parameter influence weight matrix, mapping the comprehensive score and the degradation parameter set to a high-dimensional feature space through the multidimensional projection transformation matrix to obtain a quality feature projection set; The quality feature projection set is iteratively optimized based on a support vector machine classification algorithm to obtain a quality determination result, which includes an abnormal mark and a multi-level quality classification label.

[0056] In the above embodiment, according to the statistical characteristics of the historical comprehensive score, a multi-scale sliding window is constructed, the time series of the comprehensive score is segmented and scanned and local features are extracted, the mean, variance and skewness in each window are calculated, and the abnormal preliminary distribution sequence is obtained, which reflects the local fluctuation characteristics of the comprehensive score.

[0057] According to the statistical distribution characteristics of the abnormal preliminary distribution sequence, combined with the adaptive threshold algorithm, the dynamic anomaly detection threshold is calculated, the comprehensive score is filtered point by point, and the abnormal points that exceed the threshold are marked to obtain an abnormal mark set, which includes the index of the abnormal point and the abnormal degree value. According to the distribution density and abnormal degree value of the abnormal points in the abnormal mark set, the abnormal points are grouped and clustered by the weighted K-means clustering algorithm, and the center position and weighted feature distance of each cluster are calculated to obtain the abnormal cluster feature vector, which represents the spatial aggregation pattern of the abnormal points.

[0058] According to the cluster distribution characteristics of the abnormal clustering feature vector, a mapping relationship model between the degradation parameter set and the comprehensive score is constructed. The partial least squares regression algorithm is used to decompose the contribution of each parameter in the degradation parameter set, and the influence weight of each parameter on the abnormal clustering is calculated to obtain a parameter influence weight matrix, which reflects the relative contribution of each degradation parameter to the abnormality. According to the weight distribution of the parameter influence weight matrix, a multidimensional projection transformation matrix is ​​constructed, and the comprehensive score and the degradation parameter set are mapped to a high-dimensional feature space. The feature distance and distribution density after projection are calculated to obtain a quality feature projection set, which represents the joint quality characteristics of the comprehensive score and the degradation parameter.

[0059] According to the distribution characteristics of the quality feature projection set, the support vector machine classification algorithm (SVM) is used to divide the quality level by iteratively optimizing the hyperplane, combining kernel function mapping and soft interval adjustment, calculating the classification boundary and dividing the quality interval to obtain the quality judgment result, which includes abnormal marks and multi-level quality classification labels. Specifically, a set of features extracted from the original data is repeatedly adjusted to distinguish different quality categories as well as possible by adjusting a hyperplane (the boundary that separates data in high-dimensional space). The classification result finally obtained by optimization is used to judge the quality of the data. If the data cannot be clearly separated by a straight line or plane in the original space, the data is mapped to a higher-dimensional space through a kernel function (such as a Gaussian kernel, radial basis function, etc.) so that they become separable in the high-dimensional space. In real data, there may be noise or outliers, which makes it impossible to completely separate linearly. Soft intervals allow some data points to be misclassified, and the optimization goals of classification error and boundary width are balanced by introducing penalty parameters (such as C parameters). In simple terms, we use support vector machines (SVM) to judge the quality. Specifically, we first get a bunch of feature data related to quality, and then use SVM to find a "dividing line" (hyperplane). By repeatedly adjusting this dividing line, it can best distinguish different quality levels. If the data is difficult to separate, use the kernel function to "pull" them into a space that is easier to separate, and will tolerate some small errors (soft intervals). Finally, calculate the dividing line, divide the quality into several intervals, and then mark the abnormalities and quality levels.

[0060] Reference Figure 2 The present invention also proposes a testing device for insulating glue, comprising: The acquisition module 100 is used to synchronously acquire multi-dimensional performance parameters of the insulating adhesive to be tested to obtain an initial data set, wherein the initial data set includes a resistivity sequence, a dielectric constant sequence, and a tensile strength sequence; An extraction module 200 is used to extract the feature vector of the initial data set, and construct a matrix based on the feature vector to obtain a performance feature matrix; The experimental module 300 is used to perform a dynamic loading experiment on the insulating adhesive to be tested based on the performance characteristic matrix to generate a dynamic response data set; A construction module 400 is used to construct a time series curve of the dynamic response data set, and perform degradation trend fitting processing on the time series curve to obtain a degradation parameter set; An evaluation module 500, configured to perform a multi-dimensional evaluation on the degradation parameter set based on information entropy to obtain a comprehensive score; The determination module 600 is used to perform anomaly detection on the comprehensive score, perform contribution analysis on the detected comprehensive score and the degradation parameter set, and iteratively divide the analysis result into quality levels to obtain a quality determination result.

[0061] In the above embodiment, by synchronously collecting the multi-dimensional performance parameters of the insulating glue to be tested, a performance characteristic matrix is ​​constructed, so that the test data can fully reflect the performance of the insulating glue in different dimensions. Based on the performance characteristic matrix, a dynamic loading experiment is carried out to generate a dynamic response data set, so that the test method not only focuses on the initial performance indicators of the insulating glue, but also can capture the response characteristics of the material under dynamic use conditions, thereby more accurately reflecting its long-term stability. A time series curve of the dynamic response data set is further constructed, and the curve is subjected to degradation trend fitting processing to form a degradation parameter set, so as to achieve quantitative analysis of the performance change trend of the insulating glue, which is helpful to accurately predict its durability. In addition, information entropy is used to perform multi-dimensional evaluation of the degradation parameter set to generate a comprehensive score, so that the quality judgment evaluates the overall quality level of the insulating glue through the fusion of multi-dimensional information, thereby improving the scientificity and accuracy of the test. At the same time, based on anomaly detection, the comprehensive score is analyzed, and combined with the contribution of the degradation parameter set, potential quality risks can be effectively identified to avoid misjudgment caused by abnormal local indicators. Further, by iteratively dividing the quality level, it is ensured that the quality evaluation results can more reasonably guide the selection and application of the insulating glue, and improve the reliability and service life of the insulating glue in actual engineering.

[0062] Reference Figure 3 In an embodiment of the present application, a computer device is also provided. The computer device may be a server, and its internal structure may be as follows: Figure 3 As shown. The computer device includes a processor, a memory, a network interface and a database connected through a system bus. Among them, the processor designed by the computer is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store data such as a test method for insulating glue. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a test method for insulating glue is implemented.

[0063] An embodiment of the present application also provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, a method for testing an insulating adhesive is implemented, including: synchronously collecting multi-dimensional performance parameters of the insulating adhesive to be tested to obtain an initial data set, wherein the initial data set includes a resistivity sequence, a dielectric constant sequence, and a tensile strength sequence; extracting a characteristic vector of the initial data set, and constructing a matrix based on the characteristic vector to obtain a performance characteristic matrix; performing a dynamic loading experiment on the insulating adhesive to be tested based on the performance characteristic matrix to generate a dynamic response data set; constructing a time series curve of the dynamic response data set, performing degradation trend fitting processing on the time series curve to obtain a degradation parameter set; performing a multi-dimensional evaluation on the degradation parameter set based on information entropy to obtain a comprehensive score; performing anomaly detection on the comprehensive score, performing contribution analysis on the detected comprehensive score and the degradation parameter set, and iteratively dividing the analysis results into quality grades to obtain a quality judgment result.

[0064] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media provided in this application and used in the embodiments may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double-speed data rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0065] The above description is only a preferred embodiment of the present invention, and does not limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made by using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A method for testing insulating adhesive, characterized in that: include: Synchronously collecting multi-dimensional performance parameters of the insulating adhesive to be tested to obtain an initial data set, wherein the initial data set includes a resistivity sequence, a dielectric constant sequence, and a tensile strength sequence; Extracting the characteristic vector of the initial data set, and constructing a matrix based on the characteristic vector to obtain a performance characteristic matrix; Based on the performance characteristic matrix, a dynamic loading experiment is performed on the insulating adhesive to be tested to generate a dynamic response data set; Constructing a time series curve of the dynamic response data set, performing degradation trend fitting processing on the time series curve, and obtaining a degradation parameter set; Performing a multi-dimensional evaluation on the degradation parameter set based on information entropy to obtain a comprehensive score; An abnormality detection is performed on the comprehensive score, a contribution analysis is performed on the comprehensive score after the detection and the degradation parameter set, and the analysis result is iteratively divided into quality levels to obtain a quality judgment result.

2. The method for testing insulating adhesive according to claim 1, characterized in that: The step of synchronously collecting the multi-dimensional performance parameters of the insulating adhesive to be tested to obtain an initial data set, wherein the initial data set includes a resistivity sequence, a dielectric constant sequence, and a tensile strength sequence, comprises: Perform multi-channel physical excitation on the insulating adhesive to be tested, and perform time stamp alignment processing on the excitation signal to obtain the original measurement signal; Performing signal segmentation on the original measurement signal, and performing multi-dimensional parameter conversion processing on the segmented signal to obtain preliminary performance data, wherein the preliminary performance data includes resistivity, dielectric constant and tensile strength; Performing dynamic filtering processing on the preliminary performance data to obtain a smoothed data sequence, and performing statistical analysis on the smoothed data sequence based on a sliding window to obtain a screened data set; Segmenting the screened data set into time series, and performing feature labeling processing on the segmentation results to obtain a segmented feature data set; The segmented feature data sets are integrated in the order of timestamps to obtain an initial data set.

3. The method for testing insulating adhesive according to claim 1, characterized in that: The step of extracting the characteristic vector of the initial data set and constructing a matrix based on the characteristic vector to obtain a performance characteristic matrix includes: Performing multi-scale hierarchical processing on the initial data set based on the resistivity sequence, the dielectric constant sequence and the tensile strength sequence to obtain a hierarchical sequence group; Performing nonlinear mapping on the hierarchical sequence group to obtain a primary feature vector set, and performing weighted fusion processing on the hierarchical sequence group based on the primary feature vector set to obtain a fusion feature matrix; Performing singular value decomposition processing on the fused feature matrix to obtain a reduced-dimensional feature matrix; Based on the projection value of each dimension in the dimension-reduced feature matrix, the primary feature vector set is subjected to residual correction processing to obtain an optimized feature vector set; The optimized feature vector set is subjected to tensor reconstruction processing to obtain a performance feature matrix.

4. The method for testing insulating adhesive according to claim 1, characterized in that: The step of performing a dynamic loading experiment on the insulating adhesive to be tested based on the performance characteristic matrix to generate a dynamic response data set includes: Performing loading parameter mapping processing on the performance characteristic matrix to obtain a loading condition sequence; Performing a progressive multi-field loading experiment on the insulating adhesive to be tested according to the loading condition sequence to obtain an initial response data set; Performing multi-scale time-frequency decomposition processing on the initial response data set to obtain a time-frequency feature component set; Performing adaptive noise reduction processing on the initial response data set according to the time-frequency feature component set to obtain a smoothed response data set; Performing state transition probability modeling on the smooth response data set, constructing a state transition relationship between parameters in the smooth response data set, and obtaining a dynamic evolution feature matrix; The smooth response data set is subjected to multi-parameter fusion processing according to the dynamic evolution characteristic matrix to obtain a dynamic response data set.

5. The method for testing insulating adhesive according to claim 1, characterized in that: The step of constructing a time series curve of the dynamic response data set, performing degradation trend fitting processing on the time series curve, and obtaining a degradation parameter set includes: Performing multi-dimensional time series decomposition processing on the dynamic response data set to obtain a primary time series group; Performing curve smoothing processing on the primary time series group, and marking the smoothing results in sections to obtain a section trend series set; Setting the initialization decay rate of each trend segment in the segmented trend sequence set, iteratively optimizing each of the initialization decay rates based on an optimization algorithm, and obtaining a fitting parameter initial value set; Performing dynamic weight fitting processing on the segmented trend sequence set according to the initial value set of the fitting parameters to obtain an optimized time series curve; Extracting degradation features of the optimized time series curve, and reorganizing the degradation features according to performance dimensions to obtain a degradation feature matrix; A multi-scale residual analysis is performed on the degradation feature matrix, and the degradation feature matrix is ​​optimized based on the distribution characteristics of the residual results to obtain a degradation parameter set.

6. The method for testing insulating adhesive according to claim 1, characterized in that: The step of performing multi-dimensional evaluation on the degradation parameter set based on information entropy to obtain a comprehensive score comprises: Performing single feature extraction on the degradation parameter set to obtain a single performance indicator set, and performing probability distribution calculation on information entropy based on the single performance indicator set to obtain an entropy value sequence; Setting an initial weight vector, and performing nonlinear mapping processing on the initial weight vector according to each entropy value in the entropy value sequence to obtain an optimized weight set; Performing weighted fusion processing on the individual performance indicator sets according to the optimization weight set to obtain a preliminary evaluation score; Performing residual analysis on the preliminary evaluation score to obtain an optimized evaluation score; A comprehensive characterization function is constructed based on the optimization evaluation score, and the optimization evaluation score is subjected to comprehensive characterization processing according to the comprehensive characterization function to obtain a comprehensive score.

7. The method for testing insulating adhesive according to claim 1, characterized in that: The step of performing abnormality detection on the comprehensive score, performing contribution analysis on the detected comprehensive score and the degradation parameter set, and iteratively dividing the analysis result into quality grades to obtain a quality determination result includes: Performing multi-scale sliding window processing on the comprehensive score to obtain a preliminary abnormal distribution sequence; According to the statistical distribution characteristics of the abnormal preliminary distribution sequence, dynamic anomaly detection is performed on the comprehensive score to obtain an abnormal label set; According to the distribution density and abnormal degree of the abnormal points in the abnormal mark set, weighted cluster analysis is performed on the abnormal points in the abnormal mark set to obtain an abnormal cluster feature vector; Decomposing the degradation parameter set according to the abnormal clustering feature vector to obtain a parameter influence weight matrix; Constructing a multidimensional projection transformation matrix based on the parameter influence weight matrix, mapping the comprehensive score and the degradation parameter set to a high-dimensional feature space through the multidimensional projection transformation matrix to obtain a quality feature projection set; The quality feature projection set is iteratively optimized based on a support vector machine classification algorithm to obtain a quality determination result, which includes an abnormal mark and a multi-level quality classification label.

8. A testing device for insulating glue, characterized in that: include: An acquisition module is used to synchronously acquire multi-dimensional performance parameters of the insulating adhesive to be tested to obtain an initial data set, wherein the initial data set includes a resistivity sequence, a dielectric constant sequence, and a tensile strength sequence; An extraction module, used to extract the feature vector of the initial data set, and construct a matrix based on the feature vector to obtain a performance feature matrix; An experimental module, used for performing a dynamic loading experiment on the insulating adhesive to be tested based on the performance characteristic matrix to generate a dynamic response data set; A construction module is used to construct a time series curve of the dynamic response data set, and perform degradation trend fitting processing on the time series curve to obtain a degradation parameter set; An evaluation module, used for performing a multi-dimensional evaluation on the degradation parameter set based on information entropy to obtain a comprehensive score; The determination module is used to perform anomaly detection on the comprehensive score, perform contribution analysis on the detected comprehensive score and the degradation parameter set, and iteratively divide the analysis result into quality levels to obtain a quality determination result.

9. A computer device, characterized in that: The method comprises a processor, a memory and a computer program stored in the memory and running on the processor, and the processor implements the method according to any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images