A method for testing fluorine content in fluororubber

Through the Levenberg-Marquardt fitting algorithm and LS-SVM model, the problems of long detection time and poor repetition of fluorine rubber fluorine content are solved, and accurate and stable fluorine content detection is achieved, avoiding environmental pollution.

CN111795989BActive Publication Date: 2024-11-22SHANGHAI NIUMAI ELECTRONICS TECH
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN201910279010.6
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2019-04-08
Publication Date
2024-11-22
Estimated Expiration
2039-04-08

AI Technical Summary

Technical Problem

In the prior art, the fluorine content of fluorine rubber has a long detection time, poor repeatability and environmental pollution. The existing methods are not suitable for testing of fluorine-containing rubber.

Method used

The NMR signal and time were fitted by the Levenberg-Marquardt fitting algorithm, combined with the least squares support vector machine analysis calibration model (LS-SVM), and the NMR signal-time fitting formula of the standard sample and the sample to be tested , predict the mass of fluorine element and calculate the fluorine content of fluorine rubber.

Benefits of technology

The accurate and stable detection results of fluorine rubber content are achieved, which reduces the detection time, improves the repeatability, and avoids environmental pollution.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN111795989B_ABST
    Figure CN111795989B_ABST
Patent Text Reader

Abstract

The invention belongs to the technical field of rubber detection, and particularly relates to a method for testing the fluorine content of fluororubber. The method comprises preparing a standard sample and a sample to be tested, fitting a nuclear magnetic resonance signal and time by using a Levenberg-Marquardt fitting algorithm, obtaining a nuclear magnetic resonance signal-time fitting formula, and then obtaining zero-time nuclear magnetic resonance signals of the standard sample and the sample to be tested, then establishing a least squares support vector machine analysis and calibration model (LS-SVM) between the mass of fluorine element in the standard sample and the zero-time nuclear magnetic resonance signal, predicting and analyzing the mass of fluorine element in the sample to be tested according to the analysis and calibration model of the standard sample, and obtaining the mass of fluorine element in the fluororubber. The test result obtained by fitting by the method is more accurate, and has good stability and repeatability.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention belongs to the technical field of rubber detection, and in particular relates to a method for testing the fluorine content of fluororubber. Background Art

[0002] Fluororubber is a synthetic elastomer with highly electronegative fluorine atoms on the main chain or side chain. The bond energy of the carbon-fluorine bond is very large, and the covalent radius of the fluorine atom is equivalent to half the length of the carbon-carbon bond, which can effectively protect the stability of the carbon-carbon bond. This gives it excellent properties that other rubbers cannot match, such as oil resistance, acid resistance, good physical and mechanical properties, electrical insulation and radiation resistance. It has been widely used in aerospace, automobile, petroleum, home appliances and other fields, and is an irreplaceable key material in the cutting-edge defense industry.

[0003] According to GB / T5576, fluororubber is mainly divided into three categories, namely FKM, FEPM, and FFKM. Among them, FKM refers to fluororubber containing fluorine, perfluoroalkyl or perfluoroalkoxy substituents in the polymer chain, FEPM refers to the copolymer of tetrafluoroethylene and propylene, and FFKM refers to perfluororubber with all substituents of fluorine, perfluoroalkyl or perfluoroalkoxy in the polymer. The fluorine content of these three types of rubber is mostly between 50% and 80%, and the fluorine content in fluororubber is closely related to the mechanical and chemical properties of fluororubber.

[0004] However, the national standards related to fluororubber, such as GB / T 30308-2013 and GB / T5576, do not specify the specific test method for the fluorine content. At present, the common test method for the fluorine content of solid substances is generally to subject the samples to high-temperature ashing treatment and then use ion selective electrodes or ion chromatography for testing. This is a traditional chemically destructive detection method with the disadvantages of long test time, poor repeatability, and environmental pollution during the test process.

[0005] The journal article "Measurement Method of Fluoride Content in Toothpaste Based on Nuclear Magnetic Resonance Technology" published by Yu Jie et al. discloses a method of using nuclear magnetic resonance technology to measure the fluoride signal of a standard sample. The method uses the least squares method to linearly fit the linear fitting curve of the standard sample and the signal, and then tests the signal of the fluororubber to obtain the fluorine content. However, this measurement method uses the least squares method for linear fitting, and the actual measurement result obtained is too small and has poor predictability. In addition, the sample to be tested needs to be configured into a solution during the test, which is not suitable for testing the fluorine content in fluororubber. Summary of the invention

[0006] Therefore, the technical problem to be solved by the present invention is to overcome the defects of the prior art in that the fluorine content of fluorine-containing rubber is long in detection time, has poor repeatability and stability, and causes environmental pollution during the testing process, thereby providing a nuclear magnetic resonance testing method suitable for the fluorine content in fluorine rubber with accurate detection results and good stability.

[0007] To this end, the present invention provides the following technical solutions:

[0008] The present invention provides a method for testing the fluorine content of fluororubber, comprising the following steps:

[0009] Take the fluororubber sample with known fluorine content as the standard sample, cut the standard sample and the sample to be tested into blocks respectively, and accurately weigh the mass of each sample;

[0010] The nuclear magnetic resonance signals of the standard sample and the sample to be tested are tested respectively, and the nuclear magnetic resonance signals and sampling time are fitted using the Levenberg-Marquardt fitting algorithm to obtain the nuclear magnetic resonance signal-time fitting formulas of the standard sample and the sample to be tested respectively;

[0011] According to the nuclear magnetic resonance signal-time fitting formula of the standard sample, the zero-time nuclear magnetic resonance signal of the standard sample is obtained, and then a least squares support vector machine analysis and calibration model (LS-SVM) between the fluorine element mass of the standard sample and the zero-time nuclear magnetic resonance signal is established;

[0012] According to the nuclear magnetic resonance signal-time fitting formula of the sample to be tested, the zero-time nuclear magnetic resonance signal of the sample to be tested is obtained, and then the mass of fluorine element in the sample to be tested is predicted and analyzed according to the analytical calibration model of the standard sample to obtain the mass of fluorine element in the fluororubber, thereby obtaining the fluorine content of the fluororubber.

[0013] The Levenberg-Marquardt fitting algorithm is used to fit the nuclear magnetic resonance signal and time, specifically,

[0014] The first x NMR signals of each standard sample and the corresponding sampling time are selected for data fitting. The fitting formula is:

[0015] f(t)=a1*exp(b1*t)+a2*exp(b2*t) 1.375 +a3*exp((t-b3) / c3) 2 Formula I

[0016] Where t is the sampling time; f is the NMR signal of the standard sample at time t; a1, b1, a2, b2, a3, b3, c3 are regression coefficients, and the fitting process is as follows:

[0017] (1) Prepare the time series of data DataT(t1, t2…tn) and the nuclear magnetic signal sequence DataD(d1, d2…dn):

[0018] DataD = f(t) = a1*exp(b1*t) + a2*exp(b2*t) 1.375 + a3*exp((t - b3) / c3) 2

[0019] Through the known observation vector DataD and time series vector DataT, in the data space [a, b] to be optimized, find the optimal combination of a1, b1, a2, b2, a3, b3, c3, so that after substituting the time series vector, the obtained fitting vector f(t i ) and the observation vector DataD have the smallest difference, that is, the sum of the squares of the absolute errors between the fitting vector and the observation vector is the smallest, that is, by optimizing the data combination of a1, b1, a2, b2, a3, b3, c3, obtain min(V), where n represents the dimension of the observation vector,

[0020]

[0021] (2) Initialize the algorithm parameters: the data space [a b] to be optimized, the training error ξ = 0.00001, the constant

[0022] μ 0= 0.001 and β(0 < β < 1), the vector or weight and threshold vector w to be optimized

[0023] (a < min(w) < max(w) < b) Initialize k = 0;

[0024] (3) Calculate the network output and the error index function E(w k ),

[0025]

[0026] D k = the objective function value, f’ k = the actual function value, n is the number of sample data, w is the vector composed of network weights and thresholds, e 2 k (w) is the error;

[0027] (4) Calculate the jacobian matrix of the error function:

[0028]

[0029] (5) Calculate the weight increment function according to the jacobian matrix:

[0030] Δw =[J T (w)J(w)+μI] -1 J T (w)e(w)

[0031] Among them, w k Represents the vector composed of the weights and thresholds of the kth iteration, and the vector composed of the new weights and thresholds w k+1 =w k +Δ w ;

[0032] (6) If Δ w <ξ, then go to step (8);

[0033] (7) If Δ w ≥ξ, with w k+1 =w k +Δw is the weight and threshold vector, and the error index function is calculated.

[0034] E(w k ), if E(w k+1 ) <E(w k ), then k=k+1, μ=μβ, and then go to step (3);

[0035] If E(w k+1 )≥E(w k ), then let μ = μ / β; then go to step (5) and continue the calculation;

[0036] (8) After the algorithm optimization is completed, the final result w is obtained, and the parameter combination to be optimized is converted into a1, b1, a2, b2, a3, b3, c3;

[0037] The first x NMR signals and corresponding time of each sample to be tested are selected for data fitting. The fitting formula is:

[0038] f'(t)=a4*exp(b4*t)+a5*exp(b5*t) 1.375 +a6*exp((t-b6) / c6) 2

[0039] Among them, a4, b4, a5, b5, a6, b6, c6 are regression coefficients, and the fitting process is the same as that of a1, b1, a2, b2, a3, b3, c3; t is the sampling time; f' is the nuclear magnetic resonance signal of the sample to be tested at time t.

[0040] The least squares support vector machine analysis and calibration model (LS-SVM) for establishing the relationship between the mass of the fluorine element in the standard sample and the nuclear magnetic resonance signal at zero time is specifically as follows:

[0041] Use LS-SVMMlab v1.8 toolbox, select traninlssvm function to establish analysis and calibration model, model = trainlssvm ({Y, M f Formula II

[0042] Among them, M f is the mass of fluorine element; Y is the nuclear magnetic resonance signal of the standard sample at time zero obtained by fitting; model is the analysis and calibration model.

[0043] Furthermore, when the fluorine mass of the sample to be tested is predicted and analyzed according to the analytical calibration model of the standard sample, the simlssvm function and the analytical calibration model model of formula II are selected to analyze the fluorine mass of the sample to be tested, and the model is as follows:

[0044] M f '=simlssvm(model, Y') Formula III

[0045] Among them, M f ' is the mass of fluorine element in the sample to be tested; Y' is the nuclear magnetic resonance signal of the sample to be tested at zero time obtained by fitting.

[0046] Furthermore, the x is not less than 20.

[0047] The traninlssvm function is a function obtained by optimizing the kernel function parameter δ and the regularization parameter γ using the hybrid frog leaping bionic intelligent optimization algorithm. The specific optimization process is as follows:

[0048] (1) The zero-time signal Y of the standard sample fluororubber is obtained by fitting the formula: F (Y F1 , Y F2 , Y F3 …Y Fn ) and the content of fluorine element C F (C F1 , C F2 , C F3 …C Fn );

[0049] (2) Initialize frog population parameters: frog data (N), number of meme groups (N 1 ), the number of evolutions within the meme group (N 3 ), the maximum distance of each frog jump (D), the number of population evolutions I ter ;

[0050] (3) According to the initialization parameters, randomly generate parameters: N, N 1 、N 3 ,D,I ter, where N represents the generation of N sets of parameter combinations to be optimized (kernel function parameter δ and regularization parameter γ), N 1 Represents the number of groups the combination is divided into, N 3 represents the number of times the parameter combination to be optimized is updated in each group, D represents the value added or subtracted from the original data for each update of the parameter combination to be optimized, and I ter Represents the number of algorithm cycles;

[0051] (4) confirming the number of initialization parameter combinations in each meme group;

[0052] (5) Substitute the initialization parameter combination number in each meme group into the LS-SVM model in step (1) to obtain the analysis model, and then calculate the sample fluorine content PC predicted by the analysis model F (PC F1 、PC F2 、PC F3 …PC Fn ), and finally obtain the fitness function fitness(n) corresponding to the meme group parameters:

[0053]

[0054] Among them, the smaller the fitness is, the smaller the deviation between the predicted value and the actual value is, that is, the better the fitness is;

[0055] (6) The combination of parameters to be optimized in the meme group is updated according to the following formula N 3 After each update, if the fitness is better than the previous parameter combination, the previous parameter combination will be replaced; if the fitness is worse than the previous one, it will remain unchanged.

[0056] D i =rand()*(F b -F w )

[0057] F w =F w +D i (D>D i >-D)

[0058] Among them, D i Represents the updated value of the parameter to be optimized after each update; F b Represents the best combination of parameters to be optimized within the meme group; F w Represents the parameter combination to be optimized with the worst fitness within the meme group;

[0059] (7) The above steps (4)-(6) are repeated in loop I ter Second, compare the best fitness parameter combination in all meme groups and confirm it as the final result (δ and γ);

[0060] (8) The kernel function parameter δ and regularization parameter γ finally obtained by optimization are brought into the LS-SVM model to establish an analysis model between the final fluorine content and the zero-time signal quantity, which is used to predict the fluorine content of the fluororubber of the sample to be tested.

[0061] Furthermore, the number of the standard samples is no less than 5.

[0062] The detection conditions of nuclear magnetic resonance are that the working environment temperature of nuclear magnetic resonance is 15-28℃.

[0063] The standard sample and the sample to be tested are rectangular parallelepipeds with a length of 0.2 cm-0.6 cm, a width of 0.2 cm-0.6 cm, and a height of 5 cm-10 cm.

[0064] The technical solution of the present invention has the following advantages:

[0065] 1. The test method of fluorine content of fluororubber provided by the present invention comprises preparing a standard sample and a sample to be tested, fitting the nuclear magnetic resonance signal and time by using the Levenberg-Marquardt fitting algorithm, obtaining the nuclear magnetic resonance signal-time fitting formula, and then obtaining the zero-time nuclear magnetic resonance signal of the standard sample and the sample to be tested, and then establishing a least squares support vector machine analysis and calibration model (LS-SVM) between the mass of the fluorine element in the standard sample and the zero-time nuclear magnetic resonance signal, and predicting and analyzing the mass of the fluorine element in the sample to be tested according to the analysis and calibration model of the standard sample, obtaining the mass of the fluorine element in the fluororubber, and thus obtaining the fluorine content of the fluororubber. The method adopts a combination of the LM fitting algorithm and the LS-SVM, and the test result of the fluorine content of the fluororubber obtained by fitting is more accurate, with good stability and repeatability, and in the fitting process, exponential fitting is adopted, with fast decay, and the fitted signal of the nuclear magnetic resonance of the fluororubber to be tested at zero time is more accurate; the LS-SVM analysis and calibration model has accurate fitting data, is more suitable for small samples, and the test results are more stable and have good repeatability.

[0066] 2. The test method for the fluorine content of fluororubber provided by the present invention combines the LM fitting algorithm with the LS-SVM to obtain a more accurate test result of the fluorine content of fluororubber. This is because the FID signal of fluororubber follows a multi-exponential decay, and the signal after fitting and back-extrapolation to the zero time is more consistent with the theoretical value;

[0067] The least squares support vector machine analysis and calibration model (LS-SVM) was established by using the LS-SVMMlab v1.8 toolbox and selecting the traninlssvm function. This function is a built-in function of the matlab programming language with a mature method and a stable algorithm. It has completed input and output instructions, a list of called functions, and sample codes to facilitate further optimization and improvement.

[0068] The traninlssvm function is a function after optimizing the kernel function parameter δ and the regularization parameter γ, and is optimized using the hybrid frog leaping bionic intelligent optimization algorithm. The algorithm has no requirements on the individuality of the objective function, has high optimization efficiency, can solve large-scale data optimization problems in a short time, and quickly obtain the optimal solution.

[0069] 3. The method for testing the fluorine content of fluororubber provided by the present invention can eliminate interference from strong external magnetic fields by controlling the working environment temperature of nuclear magnetic resonance to 15-28°C. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0071] Figure 1 is a nuclear magnetic resonance signal-time curve diagram of the standard sample in Example 1 of the present invention;

[0072] Figure 2 It is the standard sample nuclear magnetic resonance signal-time linear fitting curve in comparative example 1 of the present invention;

[0073] Figure 3 It is a linear fitting curve of the fluorine element mass of the standard sample in comparative example 1 of the present invention and the nuclear magnetic resonance signal at zero time. DETAILED DESCRIPTION

[0074] The following examples are provided for a better understanding of the present invention, but are not intended to limit the best mode of implementation, nor to limit the content and protection scope of the present invention. Any product identical or similar to the present invention obtained by anyone under the inspiration of the present invention or by combining the features of the present invention with other prior arts shall fall within the protection scope of the present invention.

[0075] If no specific experimental steps or conditions are specified in the examples, the conventional experimental steps or conditions described in the literature in the field can be used. If no manufacturer is specified for the reagents or instruments used, they are all conventional reagent products that can be obtained commercially.

[0076] Example 1

[0077] This embodiment provides a method for testing the fluorine content of fluororubber, and the specific steps are as follows:

[0078] Preparation of standard samples and samples to be tested: Cut the fluororubber to be tested and the standard fluororubber into rectangular samples with a length of 0.5 mm, a width of 0.5 mm, and a height of 7 cm. There are three samples to be tested, which are recorded as sample 1, sample 2, and sample 3. The fluorine content is known. The masses of the samples are 0.569 g, 0.846 g, and 0.707 g, respectively. The mass fractions of the fluorine element are 55%, 68%, and 72%, respectively. Then, the samples are placed in a quartz test tube for later use. The number of standard samples and the mass of the fluorine element are shown in Table 1.

[0079] Table 1 The mass of standard samples and the mass of fluorine element

[0080] Standard sample number Standard sample 1 Standard 2 Standard 3 Standard 4 Standard 5 Sample mass (g) 1 1 1 1 1 Mass of fluorine in the sample (g) 0.3 0.4 0.5 0.6 0.7

[0081] Signal-time fitting formula of standard sample NMR: put the standard sample into the probe coil of the NMR device, and use the FID sequence test to obtain the NMR signal, wherein the parameters of the NMR instrument are set as follows: magnetic field intensity 0.5T, resonance frequency 21.29MHz, magnet temperature 32.00±0.02℃, probe coil diameter 10mm; FID sequence: SF=21MHz, P1=2.4us, SW=1000KHz, TD=256, TW=2000ms, RFD=0.01ms, RG1=20db, DRG1=3, DR=1, PRG=3, NS=32; the working environment temperature outside the NMR is 25℃;

[0082] The first 21 NMR signals of the standard sample and the corresponding test time were selected for data fitting. The NMR signals and test time of the standard sample are shown in Figure 1 ; Using the Levenberg-Marquardt fitting algorithm, the fitting formulas are:

[0083] f1(t)=a11*exp(b11*t)+a21*exp(b21*t) 1.375 +a31*exp((t-b31) / c31) 2

[0084] f2(t)=a12*exp(b12*t)+a22*exp(b22*t) 1.375 +a32*exp((t-b32) / c32) 2

[0085] f3(t)=a13*exp(b13*t)+a23*exp(b23*t) 1.375 +a33*exp((t-b33) / c33) 2

[0086] f4(t)=a14*exp(b14*t)+a24*exp(b24*t) 1.375 +a34*exp((t-b34) / c34) 2

[0087] f5(t)=a15*exp(b15*t)+a25*exp(b25*t) 1.375 +a35*exp((t-b35) / c35) 2

[0088] Among them, f is the NMR signal of the standard sample at time t, dimensionless; t is the sampling time, in ms; the results of a, b and c after fitting are shown in Table 2:

[0089] Table 2 Specific values ​​of parameters a, b and c in the fitted formula

[0090]

[0091]

[0092] Analysis and calibration model of the mass of fluorine element in the standard sample and the nuclear magnetic resonance signal at time zero: According to the signal-time fitting formula of the nuclear magnetic resonance of the standard sample, when t=0, the nuclear magnetic resonance fitting signal Y of the standard sample at time zero is obtained. The specific results are shown in Table 3;

[0093] Table 3 Nuclear magnetic resonance signals of standard samples at time zero

[0094] Standard sample number Standard sample 1 Standard 2 Standard 3 Standard 4 Standard 5 Zero-time NMR signal 13371 16912 25508 31142 35799

[0095] Establish the mass M of fluorine in the standard sample f The least squares support vector machine analysis calibration model (LS-SVM) between the zero-time nuclear magnetic resonance signal Y, the LS-SVMlab v1.8 toolbox was used for modeling, and the traninlssvm function was selected to establish the calibration analysis model, as follows;

[0096] model = trainlssvm({Y, M f Formula II

[0097] Among them, the hybrid frog leaping bionic intelligent optimization algorithm is used to optimize the kernel function parameter δ and regularization parameter γ of the traninlssvm function. The specific optimization process is as follows:

[0098] (1) The zero-time signal Y of the standard sample fluororubber is obtained by fitting the formula: F (13371, 16912, 25508, 31142, 35799) and the content of fluorine element C F(0.3, 0.4, 0.5, 0.6, 0.7), the LS-SVM quantitative analysis model was established based on the initial kernel function parameter δ = 0.001 and the regularization parameter γ = 0.001;

[0099] (2) Initialize frog population parameters: frog data (N = 15), number of meme groups (N 1 =20), number of evolutions within the meme group (N 3 =10), the maximum distance of each frog jump (D = 5), the number of population evolutions I ter =3000;

[0100] (3) According to the initialization parameters: N, N 1 、N 3 ,D,I te , start the optimization loop;

[0101] (4) confirming the number of initialization parameter combinations in each meme group;

[0102] (5) The initialization parameter combination [δ 1 ,γ 1 ]、[δ 2 ,γ 2 ].....[δ 20 ,γ 20 ] is substituted into the LS-SVM model in step (1) to obtain the analysis model, and then the fluorine content PC of the sample predicted by the analysis model is calculated. F (PC F1 、PC F2 、PC F3 、PC F4 、PC F5 ), and finally obtain the fitness function of the corresponding meme group parameters, fitness(n), n = 5:

[0103]

[0104] (6) The combination of parameters to be optimized in the meme group is updated according to the following formula N 3 After each update, if the fitness is better than the previous parameter combination, the previous parameter combination will be replaced; if the fitness is worse than the previous one, it will remain unchanged.

[0105] D i =rand()*(F b -F w )

[0106] F w =F w +D i (D>D i >-D)

[0107] Among them, D i Represents the updated value of the parameter to be optimized after each update; F b Represents the best combination of parameters to be optimized within the meme group; F w Represents the parameter combination to be optimized with the worst fitness within the meme group;

[0108] (7) The above steps (4)-(6) are repeated in loop I ter = 3000 times, compare the best parameter combination in all meme groups and confirm it as the final result (δ = 0.00852 and γ = 0.00136);

[0109] (8) The kernel function parameter δ and regularization parameter γ finally obtained by optimization are brought into the LS-SVM model to establish an analysis model between the final fluorine content and the zero-time signal quantity, which is used to predict the fluorine content of the fluororubber of the sample to be tested.

[0110] Among them, M f is the mass of fluorine in the sample, in g; Y is the nuclear magnetic resonance signal of the standard sample at zero time obtained by fitting, dimensionless; model is the obtained analytical calibration model;

[0111] Signal-time fitting formula of nuclear magnetic resonance of the sample to be tested: put the sample to be tested into the probe coil of the nuclear magnetic resonance device, use FID sequence test to obtain the nuclear magnetic resonance signal, use the same test conditions as the standard sample for testing, select the first 21 nuclear magnetic resonance signals of the sample to be tested and the corresponding test time, use Levenberg-Marquardt fitting algorithm, and the fitting formulas are:

[0112] f 1 '(t)=a41*exp(b41*t)+a51*exp(b51*t) 1.375 +a61*exp((t-b61) / c61) 2

[0113] f 2 '(t)=a42*exp(b42*t)+a52*exp(b52*t) 1.375 +a62*exp((t-b62) / c62) 2

[0114] f 3 '(t)=a43*exp(b43*t)+a53*exp(b53*t) 1.375 +a63*exp((t-b63) / c63) 2

[0115] Among them, f' is the NMR signal of the sample to be tested at time t, dimensionless; t is the sampling time, in ms; the results of a, b and c after fitting are as follows:

[0116]

[0117] Analysis of the mass of fluorine element in the sample to be tested: According to the signal-time fitting formula of the nuclear magnetic resonance of the sample to be tested, when t=0, the nuclear magnetic resonance fitting signal Y' of the sample to be tested at time zero is obtained. The specific results are shown in Table 4;

[0118] Table 4 NMR signals of the samples to be tested at zero time

[0119] Sample number to be tested Sample 1 Sample 2 Sample 3 NMR signal of the sample to be tested at time zero 12907 23047 31037

[0120] Use LS-SVMlab v1.8 toolbox, select simlssvm function and model = trainlssvm ({Y, M f}) model to perform predictive analysis on the samples to be tested:

[0121] M f '= simlssvm(model, Y') Formula III

[0122] Among them, M f ' is the mass of fluorine in the sample to be tested, in g, see Table 5; model is the calibration model established in Formula II; Y' is the zero-time nuclear magnetic resonance signal of the sample to be tested obtained by fitting;

[0123] The mass fraction of fluorine in the sample to be tested is:

[0124] W=100*M f ' / M n

[0125] Among them, M n is the mass of fluororubber, in g; W is the mass percentage (%) of fluorine element in fluororubber, and the specific results are shown in Table 5;

[0126] Table 5 Fluorine content of fluorine rubber to be tested

[0127] Sample number to be tested Sample 1 Sample 2 Sample 3 Mass of sample to be tested (g) 0.569 0.864 0.707 The known mass of fluorine in the sample to be tested (g) 0.313 0.588 0.509 Known content of fluorine in the sample to be tested (wt%) 55 68 72 Test mass of fluorine in the sample to be tested (g) 0.311 0.582 0.506 Fluorine content in the sample to be tested (wt%) 54.657 68.171 72.277

[0128] Comparative Example 1

[0129] This comparative example provides a method for testing the fluorine content of fluororubber. The difference from Example 1 is that the nuclear magnetic resonance signal-time fitting algorithms of the standard sample and the sample to be tested are different. This comparative example uses the least squares method to perform linear fitting on the nuclear magnetic resonance signal-time of the standard sample and the sample to be tested to obtain the zero-time nuclear magnetic resonance signal of the standard sample and the sample to be tested, and then establishes a linear equation of the zero-time nuclear magnetic resonance signal and the mass of the fluorine element, and performs prediction analysis on the sample with the fluorine element content, as follows:

[0130] The least squares method is used to linearly fit the nuclear magnetic resonance signal and sampling time of the standard sample to obtain the nuclear magnetic resonance signal of the fluorine element in the sample at time zero. First, the nuclear magnetic resonance signal amplitude and sampling time are fitted to obtain the relationship between the nuclear magnetic resonance signal amplitude and sampling time of the fluorine element in the standard sample:

[0131] {(Tf 1-1 ,Tf 1-1 ,....Tf 1-21 ,Amp 1-1 ,Amp 1-2 ,.......Amp 1-21 )},{(Tf 2-1 ,Tf 1-1 ,....Tf 2-21 ,Amp 2-1 ,Amp 2-2 ,.......Amp 2-21 )},……{(Tf n-1 ,Tf n-1 ,....Tf n-21 ,Amp n-1, Amp n-2 ,.......Amp n-21 )}

[0132] Where n represents the number of samples, 21 represents the use of the first 21 data points of the sampling sequence for data fitting, and Tf n Represents the time series of sampling, assuming that it follows the linear regression model: y = a1*x+b1, then the purpose of fitting is to find:

[0133]

[0134] Finally get

[0135]

[0136] The final formula is y=a1*x+b1.

[0137] Where a represents the slope, b represents the intercept, x represents the time series, and y represents the NMR signal. When x=0, the intercept b represents the NMR signal of the sample at time zero. Finally, the NMR signals of all samples at time zero are {(b1, b2, b3.......bn)}.

[0138] There are 5 standard samples in total, each with a mass of 1g. The masses of fluorine are 0.3g, 0.4g, 0.5g, 0.6g, and 0.7g, respectively. The fitting formulas for the 5 standard samples are as follows: Figure 2 :

[0139] y 1 =-85210x+6620.6;

[0140] y 2 =-113277x+8913.1;

[0141] y 3 =-159149x+12235;

[0142] y 4 =-192952x+14890;

[0143] y 5 =-225231x+17458;

[0144] The nuclear magnetic resonance signals of the standard samples at time zero are: 6620.6, 8913.1, 12235, 14890, 17458;

[0145] The least squares method is used to perform linear fitting on the mass of fluorine element in the standard sample and the amplitude of the nuclear magnetic resonance signal (nuclear magnetic resonance signal at zero time), that is, the data fitting of binary data, {(0.3, 6620.6), (0.4, 8913.1), (0.5, 12235), (0.6, 14890), (0.7, 17458)}, assuming in advance that it follows the form of a linear function, y° = a2*x° + b2, then the purpose of fitting is to find:

[0146]

[0147] Finally get

[0148]

[0149] The final formula is y°=a2*x°+b2;

[0150] Among them, x° represents the NMR signal, y° represents the mass of fluorine element, a2 represents the slope, the specific value is 27652; b2 represents the intercept, the specific value is -1802.5, the fitting curve is shown in Figure 3 ;

[0151] For the samples to be tested, their masses were weighed as {0.569, 0.864, 0.707} respectively. The zero-time nuclear magnetic resonance signal of the fluororubber sample to be tested was calculated according to the above steps. The fitting results are shown in Table 6, and {7234.174, 13765.576, 11282.426} was obtained. Substituting into the above formula, the fluorine element mass of the corresponding sample was obtained as {0.327, 0.563, 0.473}. The calculation results are shown in Table 7.

[0152] Table 6 NMR signals of the samples to be tested at zero time

[0153] Sample number to be tested Sample 1 Sample 2 Sample 3 NMR signal of the sample to be tested at time zero 7234.174 13765.576 11282.426

[0154] Table 7 Fluorine content of test samples

[0155] Sample number to be tested Sample 1 Sample 2 Sample 3 Mass of sample to be tested (g) 0.569 0.864 0.707 The known mass of fluorine in the sample to be tested (g) 0.313 0.588 0.509 Known content of fluorine in the sample to be tested (wt%) 55 68 72 Test mass of fluorine in the sample to be tested (g) 0.327 0.563 0.473 Fluorine element (wt%) 57.434 65.162 66.931

[0156] The purpose of providing rubber with known fluorine content as a sample to be tested is to illustrate the accuracy of the test method provided by the present invention. As can be seen from Table 5 and Table 7, compared with the test results of Example 1, the test method for the fluorine content of fluororubber provided by the present invention is closer to the actual result, and the test method provided by the present invention is more accurate.

[0157] Obviously, the above embodiments are merely examples for the purpose of clear explanation, and are not intended to limit the implementation methods. For those skilled in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the implementation methods here. The obvious changes or modifications derived from these are still within the scope of protection of the invention.

Claims

1. A method for testing the fluorine content of fluororubber, characterized in that: The following steps are included: Take the fluororubber sample with known fluorine content as the standard sample, cut the standard sample and the sample to be tested into blocks respectively, and accurately weigh the mass of each sample; The nuclear magnetic resonance signals of the standard sample and the sample to be tested are tested respectively, and the nuclear magnetic resonance signals and sampling time are fitted using the Levenberg-Marquardt fitting algorithm to obtain the nuclear magnetic resonance signal-time fitting formulas of the standard sample and the sample to be tested respectively; According to the nuclear magnetic resonance signal-time fitting formula of the standard sample, the zero-time nuclear magnetic resonance signal of the standard sample is obtained, and then a least squares support vector machine analysis and calibration model (LS-SVM) between the fluorine element mass of the standard sample and the zero-time nuclear magnetic resonance signal is established; According to the nuclear magnetic resonance signal-time fitting formula of the sample to be tested, the zero-time nuclear magnetic resonance signal of the sample to be tested is obtained, and then the mass of the fluorine element in the sample to be tested is predicted and analyzed according to the analytical calibration model of the standard sample to obtain the mass of the fluorine element in the fluororubber, thereby obtaining the fluorine content of the fluororubber; The Levenberg-Marquardt fitting algorithm is used to fit the nuclear magnetic resonance signal and time, and the fitting process is specifically as follows: The first x NMR signals of each standard sample and the corresponding sampling time are selected for data fitting. The fitting formula is: f(t) = a1*exp(b1*t) + a2*exp(b2*t) 1.375 + a3*exp((t - b3) / c3) 2 Equation Ⅰ Wherein, a1, b1, a2, b2, a3, b3, c3 are regression coefficients; t is the sampling time; f is the NMR signal of the standard sample at time t; The first x NMR signals of each sample to be tested and the corresponding sampling time are selected for data fitting. The fitting formula is: f'(t)=a4*exp(b4*t)+a5*exp(b5*t) 1.375 +a6*exp((t-b6) / c6) 2 Wherein, a4, b4, a5, b5, a6, b6, c6 are regression coefficients; t is the sampling time; and f' is the nuclear magnetic resonance signal of the sample to be tested at time t.

2. The method according to claim 1, characterized in that The least squares support vector machine analysis and calibration model (LS-SVM) between the fluorine element mass of the standard sample and the nuclear magnetic resonance signal at zero time is established. The specific model is: Use LS-SVMMlab v1.8 toolbox, select traninlssvm function to establish analysis and calibration model, model = trainlssvm ({Y, M f Formula II Among them, M f is the mass of fluorine element; Y is the nuclear magnetic resonance signal of the standard sample at time zero obtained by fitting; model is the analysis and calibration model.

3. The method according to claim 2, characterized in that When predicting and analyzing the fluorine mass of the sample to be tested based on the analytical calibration model of the standard sample, the simlssvm function and the analytical calibration model model of formula II are selected to analyze the fluorine mass of the sample to be tested, and the model is as follows: M f ’ = simlssvm(model, Y’) Equation III in, M f ' is the mass of fluorine element in the sample to be tested; Y' is the nuclear magnetic resonance signal of the sample to be tested at zero time obtained by fitting.

4. The method according to any one of claims 1 to 3, characterized in that The x is not less than 20.

5. The method according to claim 2, characterized in that: The traninlssvm function is a function obtained by optimizing the kernel function parameter δ and the regularization parameter γ using a hybrid frog leaping bionic intelligent optimization algorithm.

6. The method according to any one of claims 1 to 3, characterized in that The number of the standard samples is no less than 5.

7. The method according to any one of claims 1 to 3, characterized in that The detection conditions of nuclear magnetic resonance are that the working environment temperature of nuclear magnetic resonance is 15-28℃.