High polymer material performance prediction method and system

Abstract

The invention relates to the technical field of life prediction, and provides a high polymer material performance prediction method and system, and the method comprises the steps: collecting a dielectric relaxation sweep frequency curve of a high polymer material, constructing a wavelet coefficient matrix, and obtaining the modal energy of a time-frequency ridge; establishing an exponential decay model, performing linear regression and iterative optimization on the exponential decay model, and obtaining a first-order reaction rate constant of the time-frequency ridge, initial modal energy of the time-frequency ridge and a covariance matrix; determining a peak frequency drift rate and a peak width change rate of the time frequency ridge according to the time frequency ridge and the dielectric relaxation sweep frequency curve, and establishing an attenuation feature vector of the time frequency ridge; and obtaining a trained life prediction model according to the attenuation feature vectors of the time-frequency ridges corresponding to different high polymer materials, and completing performance prediction of the high polymer materials according to the life prediction model. The method can improve the prediction accuracy of the residual life of the high polymer material.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of life prediction, in particular to a polymer material performance prediction method and system. BACKGROUND

[0002] Predicting the residual life of polymer materials can help us prevent equipment failures and safety problems caused by material aging in advance, optimize material use efficiency, prolong equipment operation life, and improve economic efficiency and reliability. A dielectric relaxation sweep curve is a curve used to study the dielectric properties of a substance, reflecting the polarization and relaxation process of polar molecules or ions inside the substance under the action of an external field. The residual life of the polymer material can be predicted by the shape, main peak position and amplitude of the dielectric relaxation sweep curve of the polymer material, and early signs of aging can be captured in time.

[0003] However, the dielectric response of initial aging is very weak and is easily submerged in background noise, and the existing technology often cannot accurately separate the noise, which often makes the prediction of the residual life of the polymer material inaccurate, and a method for accurately separating the weak dielectric response caused by aging from the background noise is needed. SUMMARY

[0004] The application provides a polymer material performance prediction method and system to solve the problem that noise has an impact on the dielectric relaxation sweep curve of the polymer material that has aged, resulting in inaccurate prediction of the residual life of the polymer material, and the technical scheme adopted is as follows: In a first aspect, an embodiment of the application provides a polymer material performance prediction method, which comprises the following steps: Collecting a first preset number of dielectric relaxation sweep curves of the polymer material, constructing a wavelet coefficient matrix according to the real part curves of all the dielectric relaxation sweep curves, dividing all the time-frequency ridges in the wavelet coefficient matrix, and calculating the modal energy of the time-frequency ridges; According to the modal energy of all the time-frequency ridges, selecting a kinetic model, establishing an exponential decay model, and obtaining the first-order reaction rate constant and the initial modal energy of the time-frequency ridges by linear regression and iterative optimization of the exponential decay model; According to the center frequency of each collection time read by the time-frequency ridge and the dielectric relaxation sweep curve, determining the peak frequency drift rate and the peak width change rate of the time-frequency ridge, and combining the first-order reaction rate constant and the initial modal energy to establish the decay feature vector of the time-frequency ridge; According to the decay feature vector of the time-frequency ridge corresponding to different polymer materials, obtaining a trained life prediction model, and completing the performance prediction of the polymer material according to the life prediction model.

[0005] Further, the calculation of the modal energy of the time-frequency ridge comprises the following specific steps: According to the time-frequency ridge, the integral boundary of the time-frequency ridge is determined; The integral boundary of the time-frequency ridge is taken as an integral interval, and the integral of the modulus square of the complex wavelet coefficient with respect to the frequency scale at the collection time is performed to obtain the modal energy of the time-frequency ridge.

[0006] Further, the method for determining the integral boundary of the time-frequency ridge comprises the following steps: The time-frequency ridge is subjected to cubic spline smoothing, the center frequency and the amplitude of the time-frequency ridge at each collection time are recorded, and the amplitude is reduced to the maximum value along the frequency direction The frequency scale interval corresponding to the collection time of the maximum value is taken as the integral boundary of the time-frequency ridge.

[0007] Further, the method for selecting the dynamic model according to the modal energy of all time-frequency ridges comprises the following specific steps: The AIC value of each dynamic model is calculated according to the modal energy of the time-frequency ridge, and the model with the minimum AIC value is taken as the selected dynamic model.

[0008] Further, the exponential decay model specifically comprises:

[0009] wherein, Ei (tj) represents the modal energy of the i-th time-frequency ridge at the j-th collection time; Ei (tj) represents the modal energy of the i-th time-frequency ridge at the j-th collection time; Ei (tj) represents the modal energy of the i-th time-frequency ridge at the j-th collection time; tj represents the order of the j-th collection time among all collection times; tj represents the order of the j-th collection time among all collection times; Ei represents the initial modal energy of the i-th time-frequency ridge; Ei represents the initial modal energy of the i-th time-frequency ridge; Ei represents the initial modal energy of the i-th time-frequency ridge; Ei represents the initial modal energy of the i-th time-frequency ridge; Ei represents the residual term of the collection time tj, and the embodiment assumes that the residual term is independent and identically distributed Gaussian white noise. Ei represents the initial modal energy of the i-th time-frequency ridge; Further, the method for obtaining the first-order reaction rate constant and the initial modal energy of the time-frequency ridge by performing linear regression and iterative optimization on the exponential decay model comprises the following specific method:

[0010] The values of the modal energy of all collection times are subjected to linear regression to obtain a fitting straight line, the first-order reaction rate constant is assigned as the opposite number of the slope of the fitting straight line, the initial modal energy of the time-frequency ridge is assigned as the intercept of the fitting straight line, the opposite number of the slope of the fitting straight line and the intercept of the fitting straight line are taken as initial search points, least square iterative optimization is performed to minimize the residual sum of squares, the optimal values of the first-order reaction rate constant and the initial modal energy are obtained when the iteration converges, and the first-order reaction rate constant and the initial modal energy are assigned with the optimal values. Ei represents the initial modal energy of the i-th time-frequency ridge; Ei represents the initial modal energy of the i-th time-frequency ridge;

[0011] Further, the determination method of the peak frequency drift rate and the peak width change rate of the time-frequency ridge is: The center frequency of each acquisition time is read according to the time-frequency ridge, and the center frequency is arranged in the order of the acquisition time to obtain a center frequency sequence; linear regression is performed on the center frequency sequence to obtain the slope of the fitting straight line, and the slope is recorded as the peak frequency drift rate of the time-frequency ridge; On the dielectric relaxation sweep curve, the two frequency points corresponding to half of the maximum value of all peak heights are calculated with the center frequency as the center, the frequency points are arranged in the order of the center frequency corresponding to the frequency points in the center frequency sequence to obtain a half-height full-width sequence of the time-frequency ridge; linear regression is performed on the half-height full-width sequence to obtain the slope of the fitting straight line, and the slope is recorded as the peak width change rate of the time-frequency ridge.

[0012] Further, the attenuation feature vector of the time-frequency ridge is specifically: The first-order reaction rate constant of the time-frequency ridge, the initial modal energy of the time-frequency ridge, the peak frequency drift rate and the peak width change rate are arranged in sequence to obtain the attenuation feature vector of the time-frequency ridge.

[0013] Further, the method for obtaining the trained life prediction model according to the attenuation feature vector of the time-frequency ridge corresponding to different high polymer materials, and completing the performance prediction of the high polymer material according to the life prediction model includes: The actual failure time of a preset number of high polymer materials is obtained by testing, the actual failure time of the high polymer material is taken as the label of the attenuation feature vector of the time-frequency ridge corresponding to the high polymer material, a data set is formed, the GPR Gaussian process regression model is trained using the data set, and a trained life prediction model is obtained; The attenuation feature vectors of all time-frequency ridges of the sample of the high polymer material to be predicted for the remaining life are input into the life prediction model to obtain the remaining life of the high polymer material to be predicted for the remaining life.

[0014] In a second aspect, the embodiments of the present application also provide a high polymer material performance prediction system, which comprises a memory, a processor, and a computer program stored in the memory and running on the processor, and the processor implements the steps of the method according to any one of the above aspects when executing the computer program.

[0015] The present application has the following advantages: In order to study the dielectric properties of the polymer material, the dielectric relaxation sweep curve of the polymer material is collected, and the time-frequency ridge is extracted according to the wavelet coefficient matrix of the real part curve of the dielectric relaxation sweep curve, the time-frequency ridge is the main energy channel corresponding to aging, and a time-frequency ridge is regarded as a mode, the mode and its neighborhood are integrated, the energy change caused by aging is converted into a scalar, and the mode energy is obtained; then, since the mode energy can represent the energy attenuation caused by aging, the mode energy curve is quantitatively fitted with the verified kinetic equation, the physical constant related to the oxidation reaction of the polymer material is extracted, and the first order reaction rate constant and the initial mode energy of the time-frequency ridge are obtained; when the mode energy attenuation is slow, there are different aging mechanisms, which may be that the dielectric relaxation unit corresponding to the mode itself relaxes slowly, or the polarized charge is rapidly relaxed and then polarized again, resulting in slow macroscopic attenuation, therefore, other dielectric characteristics are further introduced to distinguish different aging mechanisms, the peak frequency drift rate and the peak width change rate of the time-frequency ridge are determined, wherein the peak frequency drift rate is used to evaluate the change rate of the center relaxation frequency of the dielectric relaxation unit corresponding to the time-frequency ridge with time, the peak width change rate is used to evaluate the change rate of the relaxation time distribution width of the same polarization unit, and then the attenuation feature vector of the time-frequency ridge is established; finally, according to the attenuation feature vector of the time-frequency ridge corresponding to different polymer materials, a trained life prediction model is obtained, according to the life prediction model, the performance prediction of the polymer material is completed, the problem that the dielectric relaxation sweep curve of the polymer material subjected to aging is affected by noise is solved, the problem that the prediction of the remaining life of the polymer material is inaccurate is solved, and the accuracy of the prediction of the remaining life of the polymer material is improved. BRIEF DESCRIPTION OF DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description only constitute some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor.

[0017] Figure 1 The flowchart of the polymer material performance prediction method provided by an embodiment of the present application is shown in the figure. Figure 2 The wavelet coefficient matrix acquisition flowchart provided by an embodiment of the present application is shown in the figure. DETAILED DESCRIPTION

[0018] The technical solutions in the embodiments of the present application will be clearly and completely described with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0019] Please refer to Figure 1 , which shows a flow chart of a polymer material performance prediction method provided by an embodiment of the present application. The method comprises the following steps: Step S001, a first preset number of dielectric relaxation sweep curves of the polymer material are collected. A wavelet coefficient matrix is constructed according to a real part curve of the dielectric relaxation sweep curve. All time-frequency ridges in the wavelet coefficient matrix are divided. The modal energy of the time-frequency ridge is calculated.

[0020] The dielectric relaxation sweep curve is a curve for studying the dielectric properties of a substance. It is mainly used to describe the relationship between the dielectric constant or dielectric loss and the frequency under the action of an external alternating electric field. The dielectric relaxation sweep curve reflects the polarization and relaxation process of the internal polar molecules or ions of the substance under the action of an external field, and can be used to analyze the molecular structure, dynamic behavior and dielectric properties of the material. Common dielectric relaxation sweep curves include real part curves and imaginary part curves of complex dielectric constant.

[0021] The dielectric relaxation sweep curve of the polymer material can reflect the relaxation process of molecular motion mechanisms such as dipole orientation, interfacial polarization and ion migration at different time scales. In the aging process of the polymer material, the changes of chain segment activity, polar group concentration and microstructure will change the peak strength and peak position of the dielectric relaxation sweep curve, so the aging process of the polymer material can be reflected by the dielectric relaxation sweep curve.

[0022] The dielectric relaxation process shows typical time-frequency coupling characteristics in the sweep curve. From the frequency dimension, the dipole relaxation rate determines the speed of the dielectric constant real part curve in each frequency band. From the time dimension, the aging process determines the amplitude decay and extreme point drift of the whole curve. Therefore, according to the real part curve of the dielectric relaxation sweep curve, the weak dielectric response caused by aging can be separated from the background noise by using the adjustable time-frequency resolution of continuous wavelet transform through multi-scale analysis.

[0023] An impedance analyzer is used to collect a first preset number of dielectric relaxation sweep curves of the polymer material. Specifically, a stainless steel flange is welded on the side wall of the aging oven, and a parallel plate dielectric test fixture is installed. The fixture and the flange are sealed with a high-temperature-resistant PTFE insulation sleeve to prevent the atmosphere in the aging oven from leaking. The temperature of the aging oven is set to , and the relative humidity oxygen partial pressure ; the frequency range is set to , the number of scanning points is 512, the integration time is 1 second per point, the sampling interval is 10 minutes, and the dielectric relaxation sweep curves of 100 sampling time points are collected; the polymer material film with a size of , and the formula of the material is recorded; the polymer material film is placed between the parallel plate electrodes to ensure that it is perpendicular to the electric field direction and tightly attached; time sequence sampling is performed to obtain the dielectric relaxation sweep curves of the first preset number of polymer materials.

[0024] The first preset number is a preset constant, and the value of the first preset number in the embodiment is 100, i.e., the dielectric relaxation sweep curves of 100 sampling time points are collected.

[0025] The real part curve of each dielectric relaxation sweep curve is preprocessed, and the preprocessing includes denoising and normalization. The preprocessing of the real part curve of the dielectric relaxation sweep curve is a known technology and will not be described in detail. Specifically, the ALS asymmetric least squares algorithm is used for baseline correction of the real part curve in the embodiment, low-frequency background drift is eliminated, 7-point Savitzky-Golay second-order smoothing is used for the real part curve, high-frequency burrs caused by quantitative noise and contact jitter of the impedance analyzer are suppressed, the real part value of the dielectric constant at 1 kHz is used as a reference, the real part curve is divided by normalization, and absolute amplitude deviation caused by thickness and electrode area difference between batches is eliminated. The ALS asymmetric least squares algorithm for baseline correction is such that the regularization parameter is set to , and the asymmetric weight value is 0.05.

[0026] Morlet wavelet is used as a mother function, 30 scales are generated on a logarithmic grid of scales, so that the equivalent center frequency covers 20-2000 Hz, continuous wavelet transform CWT is performed on the real part curve, for each scale, the mother function and the real part curve are convolved to obtain a complex wavelet coefficient sequence composed of complex wavelet coefficient sequences of different frequencies at the scale, and a matrix composed of complex wavelet coefficient sequences corresponding to the real part curve of the dielectric relaxation sweep curve is denoted as a wavelet coefficient matrix.

[0027] It can be understood that the complex wavelet coefficients are all complex numbers, the amplitude of the complex wavelet coefficient represents the local amplitude energy at the corresponding frequency and the corresponding scale, and the phase of the complex wavelet coefficient represents the instantaneous phase. The wavelet coefficient matrix acquisition flow chart is shown in Figure 2 .

[0028] When polymer materials age, the real part curve changes, and the amplitude and phase of each wavelet coefficient in the complex wavelet coefficient sequence corresponding to the real part curve also change. When aging causes changes in polymer chain segment motion or interfacial polarization, the increase or decrease in dipole concentration will cause the amplitude to reach an extreme value. At the same time, the drift of the relaxation peak position will cause the phase to exhibit an approximately linear phase gradient along the scale direction. Therefore, when polymer materials age, the changes in the amplitude and phase of each wavelet coefficient in the complex wavelet coefficient sequence corresponding to the real part curve along the scale direction are not independent.

[0029] A curve with a local maximum amplitude and continuous phase in the wavelet coefficient matrix is ​​the main energy channel corresponding to aging, which is the time-frequency ridge.

[0030] Identify all time-frequency ridges in the wavelet coefficient matrix. Identifying time-frequency ridges in the wavelet coefficient matrix is ​​a well-known technique and will not be elaborated further. Specifically, at each acquisition time, calculate the first derivative of the amplitude point by point along the frequency axis. Mark the locations where the derivative is zero and the second derivative is negative as candidate maxima. Record candidate maxima points that are local maxima at three or more consecutive frequency points as ridge seed points. Take the phase of each ridge seed point and search for a phase difference less than a certain value in the frequency neighborhood of adjacent time points. The point with the largest amplitude is taken as the continuation point of the next ridge of the ridge seed point. The above matching process is repeated along the time axis in both the forward and reverse directions until no point can be found that satisfies both phase and amplitude continuity. The curve formed by the ridge seed point and all the continuation points of the next ridge of the ridge seed point is denoted as the time-frequency ridge.

[0031] Among these, noise or instrument-induced drift that does not meet phase consistency requirements, even if the amplitude is occasionally high, will be excluded due to the scattered phase distribution. Therefore, the purpose of selecting ridge seed points is to eliminate random noise spikes. The extraction of time-frequency ridges can identify the main energy channel with concentrated energy and continuous phase caused by aging, while eliminating amplitude variations caused by background noise and instrument drift. This provides a data basis for subsequent energy calculations and for extracting signal features with high signal-to-noise ratio.

[0032] A time-frequency ridge is a frequency curve that varies with time. There may be multiple time-frequency ridges on a single time-frequency surface. It is necessary to further extract the energy decay channels caused by aging based on the time-frequency ridges. These energy decay channels are either relaxation mechanisms within the polymer material or independent evolution paths at different aging stages. Therefore, a time-frequency ridge is treated as a mode, and the energy change caused by aging is converted into a scalar by integrating this mode and its neighborhood to obtain the mode energy.

[0033] The time-frequency ridge is smoothed using cubic splines. The center frequency, amplitude, and phase of the smoothed time-frequency ridge at each acquisition time are recorded. The amplitude is then reduced to its maximum value along the frequency direction. The frequency scale interval corresponding to the acquisition time is denoted as the integral boundary of the time-frequency ridge.

[0034] Using the integration boundary of the time-frequency ridge as the integration interval, the summation of the squared modulus of the complex wavelet coefficients with the frequency scale at each acquisition time is calculated, and the obtained frequency integration result is denoted as the modal energy of the time-frequency ridge.

[0035] It is understandable that when the first When a time-frequency ridge exists at any of the acquisition times, the modal energy of the time-frequency ridge is denoted as the time-frequency ridge at the 1st acquisition time. Modal energy at each acquisition time.

[0036] Thus, the modal energy of all time-frequency ridges in the wavelet coefficient matrix is ​​obtained.

[0037] Step S002: Based on the modal energies of all time-frequency ridges, select a dynamic model, establish an exponential decay model, and obtain the first-order reaction rate constant and initial modal energy of the time-frequency ridges by performing linear regression and iterative optimization on the exponential decay model.

[0038] Modal energies obtained from different batches of polymer materials under different aging conditions can vary significantly, making direct performance comparisons based on modal energies impossible. Modal energy can represent the energy decay caused by aging. By quantitatively fitting the modal energy curves to validated kinetic equations, physical constants related to the oxidation reaction of polymer materials can be extracted, and a model for predicting the remaining lifetime of polymer materials can be established, enabling performance prediction.

[0039] First, based on the modal energies of all time-frequency ridges, and according to the AIC value of each dynamic model, the model with the smallest AIC value is selected as the dynamic model.

[0040] Here, AIC represents the Akaike Information Criterion. In practical applications, as another implementation method, the model with the smallest BIC value can also be selected as the dynamic model based on the BIC value of each dynamic model. BIC represents the Bayesian Information Criterion. Selecting a dynamic model based on AIC and BIC values ​​is a well-known technique and will not be elaborated further.

[0041] This application selects the exponential decay model as the kinetic model for analysis. The exponential decay model corresponds to first-order kinetics, that is, during dielectric relaxation, the decay rate of dipole concentration or interfacial polarization intensity over time is proportional to the instantaneous residual amount. The specific exponential decay model is as follows:

[0042] in, Indicates the first The time-frequency ridge in the first Modal energy at each acquisition time; Indicates the first an order of the collection time among all collection times; denotes an initial modal energy of the time-frequency ridge; denotes an initial modal energy of the time-frequency ridge; denotes a natural constant; denotes a first-order reaction rate constant; denotes a residual term of the collection time , and the embodiment assumes that the residual term is independent and identically distributed Gaussian white noise.

[0043] wherein the first-order reaction rate constant is used to reflect the speed of dipole relaxation or interfacial polarization decay, and the greater the first-order reaction rate constant, the faster the dipole relaxation or interfacial polarization decay; and the initial modal energy is used to quantitatively describe the inventory of the corresponding dielectric active unit corresponding to the corresponding modal at the beginning of aging.

[0044] linear regression is performed on the numerical values of the modal energies of all collection times to obtain a fitting straight line, the first-order reaction rate constant is assigned a value of the negative of the slope of the fitting straight line, the initial modal energy of the time-frequency ridge is assigned a value of the intercept of the fitting straight line, the negative of the slope of the fitting straight line and the intercept of the fitting straight line are used as initial search points of the Levenberg-Marquardt algorithm to reduce the number of iterations, the Levenberg-Marquardt algorithm is used for least squares iterative optimization to minimize the residual sum of squares, and meanwhile, the constraint condition is set as that the initial modal energy of the time-frequency ridge and the first-order reaction rate constant are both greater than 0, the Jacobian matrix is calculated and the parameters are updated during the iteration process, and when the residual decreases by less than or the number of iterations is greater than 500, it is judged that the iteration has reached convergence, and the optimal values of the first-order reaction rate constant, the initial modal energy of the time-frequency ridge and the covariance matrix are outputted, and the first-order reaction rate constant and the initial modal energy of the time-frequency ridge are assigned the optimal values.

[0045] wherein the base of the logarithm of the logarithm of the modal energy is the natural constant.

[0046] Autocorrelation and Q-Q plots of the residual are drawn, and if the autocorrelation is within the 95% confidence interval and the Q-Q is approximately a straight line, the assumption is valid, otherwise, other kinetic models are selected for re-fitting.

[0047] The 95% confidence interval of the first-order reaction rate constant and the initial modal energy of the time-frequency ridge is calculated using the covariance matrix or 1000 times of bootstrap resampling, and if the relative uncertainty of the first-order reaction rate constant is less than 10%, it indicates that the optimal values of the first-order reaction rate constant and the initial modal energy of the time-frequency ridge are credible, otherwise, the modal energy used in the fitting process for obtaining the optimal values is discarded.

[0048] The primary reaction rate constant and the initial modal energy of the time-frequency ridge can reflect the speed of the modal energy decreasing over time.

[0049] At this point, the primary reaction rate constant of the time-frequency ridge, the initial modal energy of the time-frequency ridge, and the covariance matrix are obtained.

[0050] In step S003, the peak frequency drift rate and the peak width change rate of the time-frequency ridge are determined according to the center frequency of each acquisition time read by the time-frequency ridge and the dielectric relaxation sweep curve, and the primary reaction rate constant and the initial modal energy are combined to establish the decay characteristic vector of the time-frequency ridge.

[0051] When the modal energy decays slowly, there are different aging mechanisms, which may be that the dielectric relaxation unit corresponding to the modal itself relaxes slowly, or that the polarization charge is quickly relaxed and then re-polarized, resulting in slow macroscopic decay. Therefore, other dielectric characteristics need to be introduced to distinguish different aging mechanisms.

[0052] The center frequency of each acquisition time is read according to the time-frequency ridge, and the center frequencies are arranged in the order of acquisition time to obtain a center frequency sequence. Linear regression is performed on the center frequency sequence to obtain the slope of the fitting straight line, which is recorded as the peak frequency drift rate of the time-frequency ridge.

[0053] The peak frequency drift rate is used to evaluate the rate of change of the center relaxation frequency of the dielectric relaxation unit corresponding to the time-frequency ridge over time. When the peak frequency drift rate is greater than zero, the center frequency moves to high frequency, indicating that the effective relaxation time of the polarization unit is shortened, which usually corresponds to an increase in crosslinking density or an increase in polar group concentration, and the system tends to harden. The polarization unit is a dipole or an interface charge. Conversely, when the peak frequency drift rate is less than zero, the center frequency moves to low frequency, indicating that the effective relaxation time is extended, which usually corresponds to main chain scission, molecular weight reduction, or an increase in interface defects, and the system tends to be brittle.

[0054] On the dielectric relaxation sweep curve, the two frequency points corresponding to half of the maximum peak height are calculated with the center frequency as the center. The frequency points are arranged in the order of the center frequency in the center frequency sequence to obtain a half-height full-width sequence of the time-frequency ridge. Linear regression is performed on the half-height full-width sequence to obtain the slope of the fitting straight line, which is recorded as the peak width change rate of the time-frequency ridge.

[0055] The peak width change rate is used to evaluate the rate of change of the relaxation time distribution width of the same polarization unit. When the peak width change rate is greater than zero, it indicates that the distribution width of the relaxation time distribution of the same polarization unit is widened, indicating that the intermolecular or intramolecular disorder degree of the polymer material increases, which may be due to the increase in crosslinking network defects, chain scission ends, or interface structure unevenness. When the peak width change rate tends to zero, it indicates that the relaxation time distribution of the same polarization unit is almost unchanged, indicating that the aging path of the polymer material is single and the structure changes uniformly.

[0056] When the peak frequency drift rate is greater than zero and the peak width change rate is greater than zero, the polymer material is more likely to be dominated by cross-linking dominant hardening aging; when the peak frequency drift rate is less than zero and the peak width change rate tends to zero, the polymer material is more likely to be dominated by chain breaking dominant embrittlement aging. Among them, both aging mechanisms will cause degradation of dielectric properties, but have different effects on residual life.

[0057] The first-order reaction rate constant of the time-frequency ridge, the initial modal energy of the time-frequency ridge, the peak frequency drift rate and the peak width change rate are arranged in turn to obtain the attenuation feature vector of the time-frequency ridge.

[0058] It can be understood that there can be multiple time-frequency ridges in the wavelet coefficient matrix, and each time-frequency ridge has a corresponding attenuation feature vector.

[0059] At this point, the attenuation feature vector of the time-frequency ridge is obtained.

[0060] In step S004, the trained life prediction model is obtained according to the attenuation feature vector of the time-frequency ridge corresponding to different polymer materials, and the performance prediction of the polymer material is completed according to the life prediction model.

[0061] The traditional OIT test is performed on the polymer material until failure, and the actual failure time of the polymer material is recorded. The actual failure time of the polymer material is used as the label of the attenuation feature vector of the time-frequency ridge corresponding to the polymer material. 1000 labels corresponding to the polymer material are obtained and composed into a data set. The data set is randomly divided into a training set, a validation set and a test set according to a 7:2:1 ratio. The GPR Gaussian process regression model is trained using the data set, the kernel function is selected as the square exponential kernel determined by automatic correlation, and the trained life prediction model is obtained.

[0062] For the polymer material to be predicted, the dielectric relaxation sweep curve of the sample of the polymer material is collected, the attenuation feature vector of all time-frequency ridges of the sample is obtained, and the attenuation feature vector of all time-frequency ridges of the sample is input into the life prediction model. The life prediction model outputs the residual life of the polymer material to be predicted.

[0063] At this point, the performance prediction of the polymer material is completed.

[0064] Based on the same inventive concept as the above method, the embodiments of the present application also provide a polymer material performance prediction system, which comprises a memory, a processor and a computer program stored in the memory and running on the processor. The processor implements the steps of any one of the above polymer material performance prediction methods when executing the computer program.

[0065] The above merely describes preferred embodiments of the present application, and is not used to limit the present application, any modification, equivalent replacement, improvement, etc. made within the principle of the present application shall be included in the protection scope of the present application.

Claims

1. A method for predicting the properties of a polymeric material, characterized by, The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps:

2. The method of claim 1, wherein the polymer material is a polymer material for a display device. The method comprises the following steps: The method comprises the following steps: The method comprises the following steps:

3. The method of claim 2, wherein the step of determining the property of the polymer material is performed by a machine learning algorithm. The method comprises the following steps: The time-frequency ridge is smoothed by cubic spline, and the center frequency and amplitude of the time-frequency ridge at each acquisition time are recorded. The amplitude is reduced to the maximum value along the frequency direction The frequency scale interval corresponding to the acquisition time of the maximum value is recorded as the integral boundary of the time-frequency ridge.

4. The method of claim 1, wherein the polymer material is a polymer material for a display device. The method comprises the following steps: The method comprises the following steps:

5. The method of claim 1, wherein the polymer material is a polymer material for a display device. The method comprises the following steps: wherein, denotes the initial modal energy of the th time-frequency ridge at the th acquisition time; denotes the order of the th acquisition time among all acquisition times; denotes the initial modal energy of the th time-frequency ridge; denotes a natural constant; denotes a first-order reaction rate constant; denotes a residual term at the th acquisition time, which in this embodiment is assumed to be independent and identically distributed Gaussian white noise.

6. The method of claim 5, wherein the step of determining the property of the polymer material is performed by a neural network. The method comprises the following steps: The method comprises the following steps:

7. The method of claim 1, wherein the step of determining the property of the polymer material is performed by a neural network. The method comprises the following steps: The method comprises the following steps: The method comprises the following steps:

8. The method of claim 1, wherein the polymer material is a polymer material for a display device. The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the The first-order reaction rate constant of the time-frequency ridge, the initial modal energy of the time-frequency ridge, the peak frequency drift rate, and the peak width change rate are sequentially arranged to obtain an attenuation feature vector of the time-frequency ridge.

9. The method of claim 1, wherein the polymer material is a polymer material for a display device. The method comprises the following steps: The actual failure time of the preset number of polymer materials is obtained by testing, and the actual failure time of the polymer material is used as the label of the attenuation feature vector of the time-frequency ridge corresponding to the polymer material to form a data set. The GPR Gaussian process regression model is trained using the data set to obtain a trained life prediction model. The attenuation feature vectors of all time-frequency ridges of the sample of the polymer material whose remaining life is to be predicted are input into the life prediction model to obtain the remaining life of the polymer material whose remaining life is to be predicted.

10. A polymer material property prediction system comprising a memory, a processor, and a computer program stored in the memory and running on the processor, wherein, The processor executes the computer program to realize the steps of the method of any one of claims 1-9.

Images