A method for accelerating the initial value calculation of a data-driven double-exponential nested model

By using a data-driven, double-exponential nested model, the problem of determining the initial values ​​of regression parameters in the aging performance evaluation of polymer materials is solved, enabling rapid and accurate parameter calculation and improving the precision and efficiency of material performance evaluation.

CN118866201BActive Publication Date: 2026-11-17SHANDONG NON METALLIC MATERIAL RESEARCH INSTITUTE
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202410980679.9
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-07-22
Publication Date
2026-11-17
Estimated Expiration
2044-07-22

AI Technical Summary

Technical Problem

Existing polymer material performance prediction algorithms are mostly based on linear regression analysis, which has poor prediction results. Nonlinear regression analysis is highly dependent on the initial values ​​of parameters. Inappropriate initial values ​​can easily lead to iterative oscillations in the fitting process, making it difficult to converge or converge to local extrema, resulting in low prediction accuracy. Furthermore, no relevant research has been found on methods for calculating the initial values ​​of nonlinear regression fitting parameters for nested functions containing other parameters within the undetermined parameters.

Method used

A double-exponential nested model driven by accelerated test data is adopted. By pre-splitting and simplifying the accelerated thermal aging test data, the initial values ​​of macro and micro parameters are calculated using linear regression. The simplified aging kinetic function is then linearized to quickly determine the initial values ​​of the regression parameters of the double-exponential nested model.

Benefits of technology

It enables the rapid and accurate acquisition of precise values ​​of thermal aging kinetic model parameters, avoiding the nonlinear regression non-convergence problem caused by inappropriate initial parameter values, and improving the accuracy of material performance evaluation and the efficiency of scientific research.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure SMS_1
    Figure SMS_1
  • Figure SMS_18
    Figure SMS_18
  • Figure SMS_20
    Figure SMS_20
Patent Text Reader

Abstract

This invention discloses a method for calculating initial values ​​of a double-exponential nested model driven by accelerated testing data, belonging to the technical field of accelerated degradation test material performance evaluation. Based on accelerated thermal aging test data, a double-exponential nested model is used as a function describing the aging characteristics of the material. This model contains macro-parameters and micro-parameters, with the micro-parameters nested within the thermal aging kinetic function as an implementation of the macro-parameters. The magnitude relationship of the two power exponents is specified, and the material thermal aging kinetic function is simplified according to the aging time based on the characteristics of the double-exponential function. The accelerated testing data is pre-splitted, and the material performance data is formally split based on the material performance residual values. The simplified two aging kinetic functions are linearized, and the initial values ​​of the macro- and micro-parameters are calculated using a linear regression method. The initial values ​​of the regression parameters are input into a nonlinear regression analysis fitting software to quickly obtain the accurate values ​​of the thermal aging kinetic model parameters.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention belongs to the field of performance evaluation technology for accelerated degradation test materials, and in particular, it is a method for calculating the initial value of a double-exponential nested model driven by accelerated test data. Background Technology

[0002] With the rapid development of modern industry, polymer materials are increasingly widely used in industrial fields. However, the aging problem of polymer materials has become the biggest limitation in their application. During storage and use, polymer materials are affected by various factors, such as light, oxygen, ozone, moisture, oil and other working media, as well as temperature, stress, and strain. These factors cause chemical changes such as cross-linking and degradation of the internal molecular structure of the material, leading to the gradual degradation of various performance indicators and the gradual exhaustion of its service life. To ensure the stability of polymer materials during storage and the reliability of their long-term use, the accurate assessment of the aging performance and service life of polymer materials has become an important research topic. For example, various prediction methods have been developed for the life assessment of rubber materials, and standards such as HG3087, GB / T20028, and GB / T27800 have been established. Existing polymer material performance prediction algorithms are mostly based on linear regression analysis, which has poor prediction results. Nonlinear regression analysis prediction algorithms are highly dependent on the initial values ​​of parameters. Inappropriate initial values ​​can easily lead to iterative oscillations in the fitting process, making it difficult to converge or converge to local extrema, resulting in low prediction accuracy. In the paper "A Cyclic Search Method for Determining Initial Values ​​of Parameters for a Biexponential Curve" by Zhu Minren, published in *Mathematics in Practice and Theory* (2003-12, Vol. 33, No. 12), a cyclic search method is proposed to determine the initial values ​​of parameters when fitting a biexponential curve using the Gauss-Newton method in the least squares sense, by making full use of observed values. While this method can predict the initial values ​​of regression parameters for biexponential functions, no relevant research has been found on methods for calculating the initial values ​​of fitting parameters for nonlinear regressions of nested functions containing other parameters within the undetermined parameters. Summary of the Invention

[0003] The technical problem solved by this invention is to provide an accelerated test data-driven method for calculating the initial value of a double exponential nested model, in order to solve the problem that the initial value of a double exponential aging kinetic model with embedded micro-parameters in regression parameters is difficult to determine in nonlinear regression analysis during the process of accelerating degradation test evaluation of material performance.

[0004] To address the aforementioned technical problems, the present invention employs the following technical solution: Based on accelerated thermal aging test data, a double-exponential nested model is used as a function to describe the aging characteristics of the material. Its fitting parameters include macro-parameters and micro-parameters, with the micro-parameters nested within the thermal aging kinetic function as an implementation of the macro-parameters. The magnitude relationship between the two power exponents is specified, and the material thermal aging kinetic function is simplified according to the aging time based on the characteristics of the double-exponential function. The accelerated test data is pre-splitified, and the material performance data is formally split based on the material performance residual values. The simplified two aging kinetic functions are linearized, and the initial values ​​of the macro- and micro-parameters are calculated using a linear regression method, thereby achieving the objective of the present invention.

[0005] The present invention relates to an accelerated experimental data-driven method for calculating initial values ​​of a double-exponential nested model, characterized by the following steps: 1) For polymeric materials whose thermal aging reaction kinetics satisfy the Arrhenius equation, based on the material's thermal properties, select no fewer than three test temperatures within a range below the material's glass transition temperature to conduct accelerated thermal aging tests on the material; group the test data according to the accelerated test temperatures to obtain no fewer than three sets of test data. t i , P i The number of groups is consistent with the number of test temperatures; 2) Take the natural logarithm of each set of material performance test data, perform linear regression fitting, and obtain the slope of the fitting equation of the logarithmic values ​​of material performance with respect to aging time at all accelerated test temperatures. Take the average of the slopes to obtain k0; if k0 < 0, then let the material performance characterization value... y i = P i If k0 > 0, then let the material property characterization value be... y i =1- P i ; 3) According to the following formula, the accelerated thermal aging test temperature value T i The conversion is performed to obtain the conversion temperature value applicable to the accelerated thermal aging kinetic function;

[0006] 4) Based on the material thermal aging reaction mechanism, the following double-exponential nested model is used as a function to describe the material thermal aging kinetics. y i =(1-e)×exp(-b i × t i )+e×exp(-di × t i ) In the formula, e and b i d i For regression macro parameters; e is the weighting coefficient of the two responses before and after, 0 < e < 1; 5) Nested parameter b in the thermal aging kinetic function i d i The following formula is used to describe it; b i =b0×[exp(-bE1×X i )] d i =d0×[exp(-dE1×X i )] In the formula, b0, bE1, d0, and dE1 are the micro-parameters of the material activation energy; b i >0, d i >0; 6) Let 0 < b i <d i The critical test is determined based on the number of tests; when the aging time is greater than or equal to the aging time corresponding to the critical test, the material thermal aging kinetic function is approximately: y i =(1-e)×exp(-b i × t i When the aging time is less than the aging time corresponding to the critical test, the material thermal aging kinetic function is approximately: y i =(1-e)+e×exp(-d i × t i ); 7) Based on the determined critical test, the material performance characterization data at each accelerated test temperature are pre-splitened, taking only the data points corresponding to the critical test and its subsequent values, to obtain the material aging performance data at each accelerated test temperature after splitting. th i , yh i ); 8) Obtain the split material aging performance data, and simplify the material thermal aging kinetic function to: y i =(1-e)×exp(-b i × t i Taking the natural logarithm of both sides of the equation for this function model, we obtain the following linear model; ln( y i)=ln(1-e)-b i × t i Material aging performance data ( th i , yh i Substitute the values ​​into the above formula and perform linear regression fitting on the model to obtain the slope, intercept, and goodness of fit of the regression line. 9) Based on the intercept of the regression line, calculate the estimated value e0 of parameter e. i ; Calculate the residual value y i -(1-e0 i ), y i -(1-e0 i Data points where ) > 0 are assigned to data ( td i , yd i The remaining data points are divided into data ( tb i , yb i This completes the formal breakdown of material aging performance data for each test temperature; 10) Collect the material aging performance data at each accelerated test temperature. tb i , yb i Substitute it into ln( y i )=ln(1-e)-b i × t i Linear regression was performed on the model to obtain the slope, intercept, and goodness of fit of the regression line, and the estimated value of parameter e at the i-th experimental temperature was calculated. Parameter b i The estimated value ; 11) Obtain the material aging performance data at each accelerated test temperature. td i , yd i Substitute the estimated value of parameter e into the test temperature for each group of tests. The material thermal aging kinetic function is simplified to y i -(1- )= ×exp(-d i × t iTaking the natural logarithm of both sides of the equation for this function model, we obtain the following linear model; ln( y i -(1- ))=ln( )-d i × t i Material aging performance data ( td i , yd i Substituting into the above formula, a linear regression is performed on the model to obtain the slope, intercept, and goodness of fit of the regression line. The parameter d at the i-th experimental temperature is then calculated. i The estimated value ; 12) For each estimated value of parameter e Take the arithmetic mean to obtain the initial value of parameter e. ; b i =b0×[exp(-bE1×X i This is transformed into the linear equation ln(b0)-X. i ×bE1=ln(b i ),Will X i Substituting the values ​​into the system of linear equations and solving them simultaneously yields the initial value of the parameter b. , ; d i =d0×[exp(-dE1×X i The equation is transformed into a linear equation ln(d0)-X. i ×dE1=ln(d i ),Will X i Substituting the values ​​into the system of linear equations and solving them simultaneously yields the initial value of the parameter d. , ; At this point, the initial values ​​of all regression parameters for the double-exponential nested model of thermal aging kinetics have been quickly calculated.

[0007] Preferably, in step 2), the linear fitting of the logarithmic value of the material performance test to the aging time only distinguishes the relationship between the increase and decrease of the material performance test value and the aging time, and uniformly sets the material performance characterization value to decrease with the aging time.

[0008] Preferably, in step 6), the test corresponding to half the number of tests at each accelerated test temperature is taken as the critical test. If half the number of tests is a decimal, it is rounded up to the nearest integer as the critical test.

[0009] Preferably, in step 6), the order of the two reactions is not distinguished in the fitting of the thermal aging kinetic function. i d i They can be exchanged.

[0010] Preferably, in step 7), during the accelerated thermal aging test of the material, at each test temperature, there should be no fewer than 10 test time points and no fewer than 6 data points selected for data pre-splitting.

[0011] Preferably, in step 9), when the aging time... t i When the aging time is less than the critical test value, the material thermal aging kinetic function is approximately: y i =(1-e)+e×exp(-d i × t i ), e×exp(-d i × t i It is always a positive number, that is y i -(1-e) is always a positive number, therefore we take y i The time interval in which -(1-e)>0 holds is approximately taken as... t i The parameter d is determined within the time interval corresponding to the aging time of the critical test. i Solve for it.

[0012] Preferably, in steps 8), 10), and 11), when performing linear regression fitting, the goodness-of-fit value must be greater than 0.8.

[0013] Preferably, in step 12), b is solved. i d i The linear equation system of the microparameter is an overdetermined system of equations, and the solution is the least squares solution of the system of equations.

[0014] This invention relates to a method for calculating initial values ​​of a double-exponential nested model driven by accelerated thermal aging test data. Based on accelerated thermal aging test data, the double-exponential nested model is used as a function describing the aging characteristics of materials. Its fitting parameters include macro-parameters and micro-parameters, with the micro-parameters nested within the thermal aging kinetic function as an implementation of the macro-parameters. The relationship between the two power exponents is specified, and the thermal aging kinetic function is simplified according to the aging time based on the characteristics of the double-exponential function. The accelerated test data is pre-splitted, and the material performance data is formally split based on the material performance residual values. The simplified two aging kinetic functions are linearized, and the initial values ​​of the macro and micro parameters are calculated using linear regression. The initial values ​​of the regression parameters are input into nonlinear regression analysis software, allowing for rapid and accurate acquisition of the precise values ​​of the thermal aging kinetic model parameters. This method can help non-statistical researchers quickly select initial values ​​for the regression parameters of the double-exponential nested model, avoiding problems such as nonlinear regression non-convergence and deviation from the true values ​​caused by inappropriate initial parameter values. It lowers the technical threshold for evaluating material performance through accelerated degradation testing and improves the efficiency of scientific research. The method of this invention has clear logic, simple calculation steps, strong versatility, and is easy to implement in programming software, requiring no special data processing software. Detailed Implementation

[0015] To make the objectives, technical methods, and advantages of this invention clearer, the technical solutions in the embodiments of this invention will be described in more detail below. The described embodiments are only some, not all, of the embodiments of this invention. The embodiments described below with reference to other sources are exemplary and intended to explain this invention, and should not be construed as limiting the invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention. The embodiments of this invention will be described in detail below.

[0016] In the following examples, the subscript i represents any one of the accelerated test temperatures.

[0017] Example 1 In this embodiment, accelerated thermal aging tests were conducted on ethylene propylene diene monomer (EPDM) rubber at accelerated test temperatures of 90℃, 80℃, and 70℃. The test data were grouped according to the accelerated test temperature, and the data in each group were arranged in ascending order of test time. The thermal aging performance test data of EPDM rubber material are shown in Table 1 below.

[0018] Table 1. Test data of thermal aging properties of EPDM rubber at different accelerated test temperatures.

[0019] For each set of material performance test data, the natural logarithm was used as the dependent variable, and the corresponding aging time was used as the independent variable. A linear regression was performed to obtain the slope of the equation fitting the logarithm of the material performance test data to the aging time at all temperatures. The average of all slopes was taken to obtain k0 = 5.9072E-04, representing the material performance characterization value. y i =1- P i .

[0020] The accelerated thermal aging test temperature value T i Substituting into equation (1), we obtain the conversion temperature value X applicable to the accelerated thermal aging kinetic function. i The performance characterization data of the rubber materials obtained through the above conversion are shown in Table 2 below.

[0021]

[0022] Table 2. Characterization data of thermal aging properties of EPDM rubber at different accelerated test temperatures.

[0023] Based on the thermal aging reaction mechanism of rubber materials, the double-exponential nested model shown in formula (2) is used as a function to describe the aging kinetics of materials.

[0024]

[0025] In formula (2), e and b i d i These are the regression macro parameters. The value of e is the weighting coefficient of the two responses, where 0 < e < 1. Nested parameter b in thermal aging kinetics function i d i It is described using formula (3).

[0026]

[0027] In formula (3), b0, bE1, d0, and dE1 are micro-parameters of the material activation energy; b i >0, d i >0.

[0028] Let 0 < b i <d i The test corresponding to half the number of tests at each accelerated test temperature is taken as the critical test. If half the number of tests is a decimal, it is rounded up to the nearest integer as the critical test. When the aging time is greater than or equal to the aging time corresponding to the critical test, the material thermal aging kinetic function is approximately:

[0029] When the aging time is less than the aging time corresponding to the critical test, the material thermal aging kinetic function is approximately:

[0030] The fitting of the thermal aging kinetic function does not distinguish the order of the two reactions, b i d i They can be interchanged. Let 0 < b i <d i Calculate parameter b i For initial values, the thermal aging kinetic function omits e×exp(-d i × t i ), parameter b i The solution is too small; calculate the parameter d. i Initially, the thermal aging kinetic function will be exp(-b i × t i The value is approximately 1, and the parameter d is approximately 1. i The solution is too large, so 0 < b should be used in the calculation. i <d i It remains established.

[0031] Based on the determined critical test, the material property characterization values ​​at each accelerated test temperature are... y i Pre-splitting is performed, taking only the critical test and the corresponding data points thereafter to obtain the split data. th i , yh i ), as shown in Table 3 below.

[0032] Table 3. Pre-split data for characterizing the thermal aging properties of EPDM rubber at different accelerated test temperatures.

[0033] Taking the natural logarithm of both sides of the equation (4) for the function model, we obtain the following linear model, as shown in formula (6).

[0034]

[0035] Data ( th i , yh i Substituting into formula (6), a linear regression fit is performed on the model to obtain the slope kh of the regression line. i , intercept wh i and goodness of fit Rh i .

[0036] Let wh i=ln(1-e), and the parameter estimate e0 at the i-th test temperature is calculated using formula (6). i Substitute e0 i Calculate the residual value y i -(1-e0 i By calculation, the intercept wh of the linear regression line... 90 =-0.3178, wh 80 =-0.2523, wh 70 =-0.3104; Goodness of fit Rh 90 =0.9423, Rh 80 =0.9487, Rh 70 =0.9637, goodness-of-fit values ​​are all greater than 0.8; parameter estimate e0 90 =0.2723, e0 80 =0.2230, e0 70 =0.2668, and the performance characterization residual data obtained at each test temperature are shown in Table 4 below.

[0037] Table 4. Residual data for characterizing the thermal aging properties of EPDM rubber at different test temperatures.

[0038] When aging time t i When the value is less than the critical value, the material thermal aging kinetic function is approximately: y i =(1-e)+e×exp(-d i × t i ), e×exp(-d i × t i It is always a positive number, that is y i -(1-e) is always a positive number, therefore we take y i The time interval in which -(1-e)>0 holds is approximately taken as... t i The parameter d is determined within the aging time range corresponding to the critical test. i Solve for it.

[0039] The material aging performance data at each accelerated test temperature were formally broken down based on the residual values. y i -(1-e0 i Data points where ) > 0 are assigned to data ( td i , yd i The remaining data points are divided into data ( tb i , yb i ), the performance characterization data of the first part at each accelerated test temperature after splitting ( td i , yd i As shown in Table 5 below, the latter part of the performance characterization data ( tb i , yb i As shown in Table 6 below.

[0040] Table 5. Characterization data of the thermal aging properties of the front portion of EPDM rubber at different accelerated test temperatures.

[0041] Table 6. Characterization data of the post-heat aging properties of EPDM rubber at different accelerated test temperatures.

[0042] Take the material aging performance data after splitting at each accelerated test temperature. tb i , yb i Substitute this into formula (6) to perform linear regression fitting on the model, and obtain the slope kb of the regression line. i , intercept wb i and goodness of fit Rb i ;Let wb i =ln(1-e), to calculate the estimated value of parameter e at the i-th test temperature. , making kb i =-b i The parameter b at the i-th test temperature was calculated. i The estimated value .

[0043] The slope kb of the linear regression line was calculated. 90 =-3.1272E-04, kb 80 =-2.2389E-04, kb 70 =-7.4720E-05; intercept wb 90 =-0.3866, wb 80 =-0.2523, wb 70 =-0.3104; Goodness of fit Rb 90 =0.9713, Rb 80 =0.9487, Rb 70=0.9637, goodness-of-fit values ​​are all greater than 0.8; parameter estimates =0.3206, =0.2230, =0.2668; =3.1272E-04, =2.2389E-04, =7.4720E-05.

[0044] Take the material aging performance data after splitting at each accelerated test temperature. td i , yd i Substitute the estimated value of parameter e into the test temperature for each group of tests. The material thermal aging kinetic function (5) is simplified to formula (7).

[0045]

[0046] Taking the natural logarithm of both sides of the equation for this function model, we obtain the following linear model, as shown in formula (8).

[0047]

[0048] Data ( td i , yd i Substituting into formula (8), a linear regression fit is performed on the model to obtain the slope kd of the regression line. i , intercept wd i and goodness of fit Rd i Let kd i =-d i The parameter d at the i-th test temperature was calculated. i The estimated value .

[0049] The slope kd of the linear regression line was calculated. 90 =-0.0092, kd 80 =-0.0102, kd 70 =-0.0024; Goodness of fit Rd 90 =0.9712, Rd 80 =0.9480, Rd 70 =0.9822, goodness-of-fit values ​​are all greater than 0.8; parameter estimates =0.0092, =0.0102, =0.0024.

[0050] For each estimated value of parameter e Take the arithmetic mean to obtain the initial value of parameter e. =0.2702; b i =b0×[exp(-bE1×X i This is transformed into the linear equation ln(b0)-X. i ×bE1=ln(b i ),Will X i Substituting the values, we obtain an overdetermined system of linear equations. Solving this system simultaneously yields the initial value of the parameter b. =0.000309, =8.9502; d i =d0×[exp(-dE1×X i The equation is transformed into a linear equation ln(d0)-X. i ×dE1=ln(d i ),Will X i Substituting the values, we obtain a system of overdetermined linear equations. Solving this system simultaneously yields the initial values ​​of the parameter d. =0.01047, =8.3873.

[0051] The initial values ​​of the parameters determined above were input into the mature Russian commercial software Rubber Ageing for nonlinear fitting calculation. After 15 iterations, the fitting converged, and the precise values ​​of the regression parameters were obtained: e=0.2646, b0=0.000343, bE1=10.1307, d0=0.01090, dE1=9.1995.

[0052] Initialize arbitrary parameters =0.5, =1, =1, =1, When the initial value of 1 was input into the mature Russian commercial software Rubber Ageing for nonlinear fitting calculation, the initial value deviated significantly, and the fitting failed to converge. The initial parameter values ​​were then adjusted to be closer to the original values. =0.5, =0.001, =1, =0.001, =1 was input into the mature Russian commercial software Rubber Ageing for nonlinear fitting calculation. After 22 iterations, the fitting converged, and the precise values ​​of the regression parameters were obtained: e=0.2646, b0=0.000343, bE1=10.1307, d0=0.01090, dE1=9.1995.

[0053] This embodiment shows that when performing fitting analysis of material thermal aging characteristics, if any initial value of the parameter is given... =0.5, =1, =1, =1, =1, with a large deviation from the initial value, nonlinear fitting is prone to failure to converge; if the initial value of the parameter is given... =0.5, =0.001, =1, =0.001, =1, the initial value is relatively close, and the fitting converges after 22 iterations; the method of this invention can give the parameter initial value close to the accurate value of the parameter, and the fitting converges after 15 iterations to obtain the accurate value of the regression parameter. It takes less time and can avoid problems such as nonlinear regression non-convergence and deviation from the true value caused by inappropriate parameter initial values, thus improving the efficiency of scientific research.

[0054] Example 2 In this embodiment, accelerated thermal aging tests were conducted on natural rubber NR at accelerated test temperatures of 85℃, 75℃, and 65℃. The test data were grouped according to the accelerated test temperature, and the data in each group were arranged in ascending order of test time. The test data of thermal aging performance of NR rubber material are shown in Table 7 below.

[0055] Table 7. Test data of thermal aging properties of NR rubber at different accelerated test temperatures.

[0056] For each set of material performance test data, the natural logarithm was used as the dependent variable, and the corresponding aging time was used as the independent variable. A linear regression was performed to obtain the slope of the equation fitting the logarithm of the material performance test values ​​to the aging time at all temperatures. The average of all slopes was taken to obtain k0 = -0.0296, which represents the material performance characterization value. y i = P i .

[0057] The accelerated thermal aging test temperature value T i Substituting into formula (1), we obtain the conversion temperature value X applicable to the accelerated thermal aging kinetic function. iThe performance characterization data of the rubber materials obtained through the above conversion are shown in Table 8 below.

[0058] Table 8. Characterization data of thermal aging properties of NR rubber at different accelerated test temperatures.

[0059] Based on the thermal aging reaction mechanism of rubber materials, the double-exponential nested model shown in formula (2) is used as a function to describe the aging kinetics of materials.

[0060] Nested parameter b in thermal aging kinetics function i d i It is described using formula (3).

[0061] Let 0 < b i <d i The test corresponding to half the number of tests at each accelerated test temperature is taken as the critical test. If half the number of tests is a decimal, it is rounded up to the nearest integer as the critical test. When the aging time is greater than or equal to the aging time corresponding to the critical test, the material thermal aging kinetic function is approximately given by formula (4). When the aging time is less than the aging time corresponding to the critical test, the material thermal aging kinetic function is approximately given by formula (5).

[0062] Based on the determined critical test, the material property characterization values ​​at each accelerated test temperature are... y i Pre-splitting is performed, taking only the critical test and the corresponding data points thereafter to obtain the split data. th i , yh i ), as shown in Table 9 below.

[0063] Table 9. Pre-split data for characterizing the thermal aging properties of NR rubber at different accelerated test temperatures.

[0064] Data ( th i , yh i Substituting into model (6), a linear regression fit is performed on the model to obtain the slope kh of the regression line. i , intercept wh i and goodness of fit Rh i .

[0065] Let wh i =ln(1-e), and the parameter estimate e0 at the i-th test temperature is calculated using formula (6). i Substitute e0 i Calculate the residual value yi -(1-e0 i By calculation, the intercept wh of the linear regression line... 85 =-0.4114, wh 75 =-0.4174, wh 65 =-0.3916; Goodness of fit Rh 85 =0.9888, Rh 75 =0.9960, Rh 65 =0.9914, goodness-of-fit values ​​are all greater than 0.8; parameter estimate e0 85 =0.3373, e0 75 =0.3412, e0 65 =0.3241, and the performance characterization residual data obtained at each test temperature are shown in Table 10 below.

[0066] Table 10 Residual data for characterizing the thermal aging properties of NR rubber at different test temperatures

[0067] The material aging performance data at each accelerated test temperature were formally broken down based on the residual values. y i -(1-e0 i Data points where ) > 0 are assigned to data ( td i , yd i The remaining data points are divided into data ( tb i , yb i ), the performance characterization data of the first part at each accelerated test temperature after splitting ( td i , yd i As shown in Table 11 below, the latter part of the performance characterization data ( tb i , yb i As shown in Table 12 below.

[0068] Table 11. Characterization data of the front part of NR rubber under different accelerated test temperatures.

[0069] Table 12 Characterization data of the rear part of NR rubber under different accelerated test temperatures

[0070] Take the material aging performance data after splitting at each accelerated test temperature. tb i , yb i Substitute this into formula (6) to perform linear regression fitting on the model, and obtain the slope kb of the regression line. i , intercept wb i and goodness of fit Rb i ;Let wb i =ln(1-e), to calculate the estimated value of parameter e at the i-th test temperature. , making kb i =-b i The parameter b at the i-th test temperature was calculated. i The estimated value .

[0071] The slope kb of the linear regression line was calculated. 85 =-0.0513, kb 75 =-0.0208, kb 65 =-0.0084; intercept wb 85 =-0.3816, wb 75 =-0.3945, wb 65 =-0.3778; Goodness of fit Rb 85 =0.9943, Rb 75 =0.9951, Rb 65 =0.9900, goodness-of-fit values ​​are all greater than 0.8; parameter estimates =0.3173, =0.3260, =0.3147; =0.0513, =0.0208, =0.0084.

[0072] Take the material aging performance data after splitting at each accelerated test temperature. td i , yd i Substitute the estimated value of parameter e into the test temperature for each group of tests. , will data ( td i , yd i Substituting into formula (8), a linear regression fit is performed on the model to obtain the slope kd of the regression line. i , intercept wd i and goodness of fit Rd i Let kd i =-d i The parameter d at the i-th test temperature was calculated. iThe estimated value .

[0073] The slope kd of the linear regression line was calculated. 85 =-1.7483, kd 75 =-0.8510, kd 65 =-0.5248; Goodness of fit Rd 85 =0.9882, Rd 75 =0.9947, Rd 65 =0.9921, goodness-of-fit values ​​are all greater than 0.8; parameter estimates =1.7483, =0.8510, =0.5248.

[0074] For each estimated value of parameter e Take the arithmetic mean to obtain the initial value of parameter e. =0.3193; b i =b0×[exp(-bE1×X i This is transformed into the linear equation ln(b0)-X. i ×bE1=ln(b i ),Will X i Substituting the values, we obtain an overdetermined system of linear equations. Solving this system simultaneously yields the initial value of the parameter b. =0.065599, =10.921290; d i =d0×[exp(-dE1×X i The equation is transformed into a linear equation ln(d0)-X. i ×dE1=ln(d i ),Will X i Substituting the values, we obtain a system of overdetermined linear equations. Solving this system simultaneously yields the initial values ​​of the parameter d. =1.976761, =7.264393.

[0075] The initial values ​​of the parameters determined above were input into the mature Russian commercial software Rubber Ageing for nonlinear fitting calculation. After 7 iterations, the fitting converged, and the precise values ​​of the regression parameters were obtained: e=0.3219, b0=0.0665068, bE1=11.1218, d0=1.08025, dE1=5.54241.

[0076] Approximate the initial value of the parameter =0.5, =1, =1, =1, =1 was input into the mature Russian commercial software Rubber Ageing for nonlinear fitting calculation. After 40 iterations, the fitting converged, and the precise values ​​of the regression parameters were obtained: e=0.3219, b0=0.0665068, bE1=11.1218, d0=1.08025, dE1=5.54241.

[0077] This embodiment shows that when performing fitting analysis of material thermal aging characteristics, if initial parameter values ​​are given... =0.5, =1, =1, =1, =1, the initial value is relatively close, and the fitting converges after 40 iterations; the method of this invention can give the parameter initial value close to the accurate parameter value, and the fitting converges after 7 iterations to obtain the accurate value of the regression parameter. It takes less time and can avoid problems such as nonlinear regression non-convergence and deviation from the true value caused by inappropriate parameter initial values, thus improving the efficiency of scientific research.

Claims

1. A method for accelerating the initial value calculation of a data-driven double-exponential nested model, characterized in that, The steps are as follows: 1) For polymeric materials whose thermal aging reaction kinetics satisfy the Arrhenius equation, based on the material's thermal properties, select no fewer than three test temperatures within a range below the material's glass transition temperature to conduct accelerated thermal aging tests on the material; group the test data according to the accelerated test temperatures to obtain no fewer than three sets of test data. t i , P i The number of groups is consistent with the number of test temperatures; 2) Take the natural logarithm of each set of material performance test data, perform linear regression fitting, and obtain the slope of the fitting equation of the logarithmic values ​​of material performance with respect to aging time at all accelerated test temperatures. Take the average of the slopes to obtain k0; if k0 < 0, then let the material performance characterization value... y i = P i If k0 > 0, then let the material property characterization value be... y i =1- P i ; 3) According to the following formula, the accelerated thermal aging test temperature value T i The conversion is performed to obtain the conversion temperature value applicable to the accelerated thermal aging kinetic function; 4) Based on the material thermal aging reaction mechanism, the following double-exponential nested model is used as a function to describe the material thermal aging kinetics. y i =(1-e)×exp(-b i × t i )+e×exp(-d i × t i ) In the formula, e and b i d i For regression macro parameters; e is the weighting coefficient of the two responses before and after, 0 < e < 1; 5) Nested parameter b in the thermal aging kinetic function i d i The following formula is used to describe it; b i =b0×[exp(-bE1×X i )] d i =d0×[exp(-dE1×X i )] In the formula, b0, bE1, d0, and dE1 are the micro-parameters of the material activation energy; b i >0, d i >0; 6) Let 0 < b i <d i The critical test is determined based on the number of tests; when the aging time is greater than or equal to the aging time corresponding to the critical test, the material thermal aging kinetic function is approximately: y i =(1-e)×exp(-b i × t i When the aging time is less than the aging time corresponding to the critical test, the material thermal aging kinetic function is approximately: y i =(1-e)+e×exp(-d i × t i ); 7) Based on the determined critical test, the material performance characterization data at each accelerated test temperature are pre-splitened, taking only the data points corresponding to the critical test and its subsequent values, to obtain the material aging performance data at each accelerated test temperature after splitting. th i , yh i ); 8) Obtain the split material aging performance data, and simplify the material thermal aging kinetic function to: y i =(1-e)×exp(-b i × t i Taking the natural logarithm of both sides of the equation for this function model, we obtain the following linear model; ln( y i )=ln(1-e)-b i x t i Material aging performance data ( th i , yh i Substitute the values ​​into the above formula and perform linear regression fitting on the model to obtain the slope, intercept, and goodness of fit of the regression line. 9) Based on the intercept of the regression line, calculate the estimated value e0 of parameter e. i ; Calculate the residual value y i -(1-e0 i ), y i -(1-e0 i Data points where ) > 0 are assigned to data ( td i , yd i The remaining data points are divided into data ( tb i , yb i This completes the formal breakdown of material aging performance data for each test temperature; 10) Collect the material aging performance data at each accelerated test temperature. tb i , yb i Substitute it into ln( y i )=ln(1-e)-b i × t i Linear regression was performed on the model to obtain the slope, intercept, and goodness of fit of the regression line, and the estimated value of parameter e at the i-th experimental temperature was calculated. Parameter b i The estimated value ; 11) Obtain the material aging performance data at each accelerated test temperature. td i , yd i Substitute the estimated value of parameter e into the test temperature for each group of tests. The material thermal aging kinetic function is simplified to y i -(1- )= ×exp(-d i × t i Taking the natural logarithm of both sides of the equation for this function model, we obtain the following linear model; ln( y i -(1- ))=ln( )-d i × t i Material aging performance data ( td i , yd i Substituting into the above formula, a linear regression is performed on the model to obtain the slope, intercept, and goodness of fit of the regression line. The parameter d at the i-th experimental temperature is then calculated. i The estimated value ; 12) For each estimated value of parameter e Take the arithmetic mean to obtain the initial value of parameter e. ; b i =b0×[exp(-bE1×X i This is transformed into the linear equation ln(b0)-X. i ×bE1=ln(b i ),Will X i Substituting the values ​​into the system of linear equations and solving them simultaneously yields the initial value of the parameter b. , ; d i =d0×[exp(-dE1×X i The equation is transformed into a linear equation ln(d0)-X. i ×dE1=ln(d i ),Will X i Substituting the values ​​into the system of linear equations and solving them simultaneously yields the initial value of the parameter d. , ; At this point, the initial values ​​of all regression parameters for the double-exponential nested model of thermal aging kinetics have been quickly calculated.

2. The method for calculating initial values ​​of a double-exponential nested model driven by accelerated experimental data according to claim 1, characterized in that: In step 2), the linear fitting of the logarithmic value of the material performance test to the aging time only distinguishes the relationship between the increase and decrease of the material performance test value and the aging time, and uniformly sets the material performance characterization value to decrease with the aging time.

3. The method for calculating initial values ​​of a double-exponential nested model driven by accelerated experimental data according to claim 1, characterized in that: In step 6), the test corresponding to half the number of tests at each accelerated test temperature is taken as the critical test. If half the number of tests is a decimal, it is rounded up to the nearest integer as the critical test.

4. The method for calculating initial values ​​of a double-exponential nested model driven by accelerated experimental data according to claim 1, characterized in that: In step 6), the thermal aging kinetic function fitting does not distinguish the order of the two reactions, b i d i They can be exchanged.

5. The method for calculating initial values ​​of a double-exponential nested model driven by accelerated experimental data according to claim 1, characterized in that: In step 7), during the accelerated thermal aging test of the material, at each test temperature, there should be no fewer than 10 test time test points, and the data pre-splitting should select no fewer than 6 data points.

6. The method for calculating initial values ​​of a double-exponential nested model driven by accelerated experimental data according to claim 1, characterized in that: In step 9), when the aging time... t i When the aging time is less than the critical test value, the material thermal aging kinetic function is approximately: y i =(1-e)+e×exp(-d i × t i ), e×exp(-d i × t i It is always a positive number, that is y i -(1-e) is always a positive number, therefore we take y i The time interval in which -(1-e)>0 holds is approximately taken as... t i The parameter d is determined within the aging time range corresponding to the critical test. i Solve for it.

7. The method for calculating initial values ​​of a double-exponential nested model driven by accelerated experimental data according to claim 1, characterized in that: In steps 8), 10), and 11), when performing linear regression fitting, the goodness-of-fit value must be greater than 0.

8.

8. The method for calculating initial values ​​of a double-exponential nested model driven by accelerated experimental data according to claim 1, characterized in that: In step 12), b is solved. i d i The linear equation system of the microparameter is an overdetermined system of equations, and the solution is the least squares solution of the system of equations.

Citation Information

Patent Citations

  • Wiener accelerated degradation model construction method and system based on exponential time function

    CN111832151A

  • Method for predicting a physical characteristic of vacuum insulation panels

    EP2594924A1