A cis-trans cyclohexane diacid diethyl ester nondestructive detection method based on feature screening

Abstract

The application discloses a cis-trans cyclopropane diacid diethyl ester nondestructive detection method based on feature screening, relates to the technical field of testing or analyzing materials by means of determining the chemical or physical properties of the materials, and comprises the following steps: taking a plurality of mixtures formed by cis isomers and trans isomers of cyclopropane diacid diethyl ester respectively at a plurality of mass ratios, and taking Raman characteristic peak parameters of the plurality of mixtures; based on the Raman characteristic peak parameters of the plurality of mixtures, a plurality of Raman characteristic peak proportional type variables between cis-trans isomers are constructed; a prediction model is constructed; and based on the prediction model, the mass ratio of the cis-trans isomers of cyclopropane diacid diethyl ester is predicted; the method has the advantages of rapidness, nondestructiveness, non-separation, high recognition precision and the like, and is suitable for the fields of isomer component analysis, structure identification and precise quantitative detection.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of testing or analyzing materials by determining the chemical or physical properties of the materials, in particular to a cis-trans cyclopropane diacid diethyl ester non-destructive detection method based on feature screening. BACKGROUND

[0002] Cyclopropane diacid diethyl ester is an important organic synthesis intermediate, widely used in the fields of medicine, pesticide and polymer material. The compound exists in cis and trans two stable spatial configurations, that is, cis-trans isomers of cyclopropane diacid diethyl ester. Although both are completely consistent in molecular composition, due to the tension effect of the three-membered ring structure and the different relative orientation of the ester group, they show significant differences in reaction activity, thermal stability, stereoselectivity and other physical and chemical properties. The cis configuration has higher reaction activity due to the close spatial position of the two ester groups, and is often used to construct chiral drug precursors and high selectivity functional monomers; while the trans configuration is more suitable for the synthesis of rigid backbone structure, thermal stability material and part of agricultural compounds due to its stable conformation and large steric hindrance. Therefore, accurate identification and quantitative analysis of the proportion of the two isomers is of great significance to ensure the selectivity of the synthesis path, optimize the product performance and improve the quality control of fine chemical production.

[0003] The existing analysis of cis-trans isomers of cyclopropane diacid diethyl ester mainly relies on chromatography or mass spectrometry technology. Chromatography has good separation and quantitative ability, but usually needs to use chiral chromatographic column or complex derivatization pretreatment, which is difficult to realize rapid response and in-situ detection. Mass spectrometry is difficult to effectively distinguish cis-trans configuration due to the same molecular mass and extremely similar structure, and needs to be combined with high-efficiency separation means to realize analysis. For example, the traditional separation method often uses solvent selective crystallization technology to obtain pure trans or cis isomers by using acetonitrile or nitromethane, which is a cumbersome process, time-consuming and strongly dependent on operating conditions, and is difficult to adapt to high-throughput or real-time monitoring requirements.

[0004] Raman spectroscopy technology, as a non-destructive detection method based on molecular vibration, has good application potential in the fields of organic molecule identification, material sorting and drug analysis due to its high selectivity, high resolution and structure sensitivity. However, the difference between cis-trans isomers in Raman spectrum is often weak, and the traditional discrimination method based on single peak or subjective experience ratio is difficult to realize high-precision identification and quantification. SUMMARY

[0005] In order to solve the above problems, the present application provides a cis-trans cyclopropane diacid diethyl ester non-destructive detection method based on feature screening.

[0006] A cis-trans cyclopropane diacid diethyl ester non-destructive detection method based on feature screening, comprising the following steps:

[0007] Respectively taking a plurality of mixtures formed by the cis isomer and the trans isomer of cyclopropane diacid diethyl ester at a plurality of mass ratios, obtaining Raman characteristic peak parameters of the plurality of mixtures;

[0008] Based on the Raman characteristic peak parameters of the plurality of mixtures, a plurality of Raman characteristic peak ratio type variables between the cis isomer and the trans isomer are constructed; the Raman characteristic peak ratio type variables are used to describe the difference between the Raman characteristic peaks of the cis isomer and the trans isomer;

[0009] A prediction model is constructed, and an input of the prediction model is the Raman characteristic peak ratio type variable, and an output of the prediction model is the mass ratio of the cis isomer and the trans isomer;

[0010] Based on the prediction model, the mass ratio of the cis isomer and the trans isomer of cyclopropane diacid diethyl ester is predicted.

[0011] Description: The above method realizes accurate and efficient differentiation detection of the mass ratio of cis-trans isomers by constructing the Raman characteristic peak ratio variable between the isomers and establishing a prediction model; for the problem that the spatial configuration difference between the cis-trans isomers is small and the traditional method is difficult to distinguish, the characteristic peak ratio variable between the isomers is used to amplify the molecular response difference, and the configuration resolution capability is significantly improved; the prediction model directly outputs the mass ratio, realizes accurate prediction from the Raman characteristic difference to the mass ratio, does not need to be repeatedly calibrated, and is suitable for the fields such as pharmacy and material science which need to accurately control the configuration ratio.

[0012] Description: Since the difference in the proportion of cis and trans cyclopropane diacid diethyl ester is of great significance to guarantee the selectivity of the synthesis path, optimize the product performance and improve the quality control of fine chemical production, it is necessary to quickly distinguish the two isomers for subsequent operations.

[0013] Further, the method for obtaining the Raman characteristic peak parameters of the plurality of mixtures comprises:

[0014] The Raman characteristic peaks of the cis isomer and the trans isomer of cyclopropane diacid diethyl ester are determined respectively;

[0015] According to the Raman characteristic peaks of the cis isomer and the trans isomer, the Raman characteristic peak parameters of the plurality of mixtures are determined, and the Raman characteristic peak parameters are the Raman characteristic peak areas of the cis isomer and the trans isomer.

[0016] Description: The above method clearly defines the Raman characteristic peaks of the cis-trans isomers, and calculates the Raman characteristic peak areas, which provides data support for subsequent construction of the ratio type variable.

[0017] Further, the method for respectively determining the Raman characteristic peaks of the cis isomer and the trans isomer of diethyl cyclopropane dicarboxylate comprises:

[0018] Respectively take the cis isomer and the trans isomer of diethyl cyclopropane dicarboxylate as samples, and respectively obtain the standard Raman spectrum of the cis isomer and the standard Raman spectrum of the trans isomer; extract the Raman characteristic peaks of the cis isomer from the standard Raman spectrum of the cis isomer, and extract the Raman characteristic peaks of the trans isomer from the standard Raman spectrum of the trans isomer.

[0019] Description: The above method takes samples and obtains standard Raman spectra, which ensures the accuracy and standardization of the measured data by using samples, and the standard Raman spectra can truly reflect the characteristics of the cis-trans isomers; extracting the Raman characteristic peaks from the standard Raman spectra makes the whole process operable and systematic.

[0020] Further, the range of the plurality of mass ratios is 1-9:10, and the Gaussian fitting method is used to calculate the Raman characteristic peak area of the cis isomer and the Raman characteristic peak area of the trans isomer.

[0021] Description: The above range can ensure that the data is sufficiently representative and covers a wide range, and the Gaussian fitting method is used to calculate the Raman characteristic peak area, which can accurately fit the peak shape, effectively reduce noise interference and measurement error, and improve the accuracy and reliability of the calculation of the characteristic peak area.

[0022] Further, the Raman characteristic peak ratio variable includes an original variable and a constructed variable obtained by combining the original variable; the original variable includes a single-peak ratio of the Raman characteristic peak area, a weighted combination ratio, a logarithmic transformation, a multi-peak product ratio, a symmetry index, a ratio of a single-peak weighted sum to a multi-peak weighted sum, a combination construction of symmetric peaks and asymmetric peaks, a cross ratio of homologous vibration modes in the Raman spectrum, a logarithmic or square root conversion of the peak area.

[0023] Description: The above content lists the composition of the Raman characteristic peak ratio variable, and the ratio variable can fully reflect the complex relationship between the Raman characteristic peaks, and further capture the difference information between the cis isomer and the trans isomer.

[0024] Further, the method for obtaining the constructed variable obtained by combining the original variable comprises:

[0025] Perform full combination traversal of the original variables in two-dimensional and three-dimensional combination spaces to generate a plurality of two-dimensional combinations and a plurality of three-dimensional combinations; and take the plurality of two-dimensional combinations and the plurality of three-dimensional combinations as the constructed variables.

[0026] Description: The above method generates the construction variable by performing full combination traversal on the original variable in the two-dimensional and three-dimensional combination space, greatly expands the dimension and combination form of the variable, can mine more complex and deeper correlation information between the original variables, and helps to more fully capture the subtle differences between the cis-isomer and the trans-isomer in the Raman characteristics.

[0027] Further, the method for constructing the prediction model comprises:

[0028] correlation analysis is performed on the Raman characteristic peak area ratio type variable and the mass ratio of the plurality of cis-isomers and trans-isomers, and at least one Raman characteristic peak area ratio type variable with the highest correlation with the plurality of mass ratios is selected;

[0029] The at least one Raman characteristic peak area ratio type variable with the highest correlation with the plurality of ratios is taken as the input of the prediction model, the mass ratio of the plurality of cis-isomers and trans-isomers is taken as the output, a machine learning model is used for model training, and after the training is completed, the prediction model is constructed.

[0030] Further, the prediction model uses one of a support vector regression model, a random forest model and a neural network model.

[0031] Description: The above method first accurately selects the Raman characteristic peak variable highly correlated with the cis / trans-isomer mass ratio as the input through correlation analysis, combines the machine learning model for training, ensures the physical meaning explicitness and data representativeness of the input features, ensures the prediction accuracy and calculation efficiency through model calculation, and constructs the prediction model with high generalization ability and interpretability.

[0032] The method of the present application realizes accurate and efficient differentiation detection of the cis-trans isomer mass ratio by constructing the Raman characteristic peak ratio variable between isomers and establishing a prediction model; for the problem that the spatial configuration difference between cis-trans isomers is small and difficult to distinguish by traditional methods, the characteristic peak ratio variable between isomers is used to amplify the molecular response difference, which significantly improves the configuration resolution ability; the prediction model directly outputs the mass ratio, realizes accurate prediction from the Raman characteristic difference to the mass ratio, does not need to be calibrated repeatedly, and is suitable for the fields such as pharmacy and material science which need to accurately control the configuration ratio; specifically, through the combination construction and variable screening mechanism, the spatial configuration difference in the Raman response is significantly enhanced, realizing rapid, non-separation cis-trans isomer recognition and quantitative analysis. BRIEF DESCRIPTION OF DRAWINGS

[0033] Figure 1 is the molecular structure formula of diethyl cyclopropane diacid cis-trans isomer in the embodiment of the present application;

[0034] Figure 2is the standard Raman spectrum of cis / trans cyclopropane diacid diethyl ester in the embodiment of the application;

[0035] Figure 3 is the Raman spectrum of a cis / trans isomer mixture in the embodiment of the application;

[0036] Figure 4 is the linear regression model and equation with the optimal linear fitting in the embodiment of the application;

[0037] Figure 5 is the Raman spectrum of a mixture with an unknown mass fraction in the embodiment of the application. DETAILED DESCRIPTION

[0038] To further illustrate the manner of carrying out the application and to further substantiate the effects achieved by the application, the technical solutions of the application will be clearly and completely described below with reference to experiments.

[0039] In combination with the problems described in the background art, the application proposes a cis-trans cyclopropane diacid diethyl ester nondestructive detection method based on feature screening, constructs a combination variable of areas of multiple groups of Raman characteristic peaks, and introduces a combination space traversal and regression screening strategy, thereby achieving, for the first time, accurate differentiation of isomer configurations and mass fraction prediction without separation, significantly improving the differentiation and detection capability of Raman technology in complex cis-trans isomer systems, and providing an efficient and reliable new solution for configuration control and quality evaluation of drug intermediates and fine chemicals. The following embodiments take the cis isomer and the trans isomer of cyclopropane diacid diethyl ester as examples for detailed description.

[0040] Embodiment 1: A cis-trans cyclopropane diacid diethyl ester nondestructive detection method based on feature screening, comprising the following steps:

[0041] S1, respectively taking multiple mixtures of cis isomers and trans isomers of cyclopropane diacid diethyl ester at multiple mass ratios to obtain Raman characteristic peak parameters of the multiple mixtures; the cyclopropane diacid diethyl ester has a structure as shown in Figure 1 , including cis cyclopropane diacid diethyl ester and trans cyclopropane diacid diethyl ester;

[0042] The method for obtaining the Raman characteristic peak parameters of the multiple mixtures comprises:

[0043] S1-1, the method for respectively determining the Raman characteristic peak of the cis isomer and the Raman characteristic peak of the trans isomer of cyclopropane diacid diethyl ester comprises:

[0044] respectively, and the standard Raman spectrum of the cis isomer and the standard Raman spectrum of the trans isomer are obtained respectively; the Raman characteristic peaks of the cis isomer are extracted from the standard Raman spectrum of the cis isomer, and the Raman characteristic peaks of the trans isomer are extracted from the standard Raman spectrum of the trans isomer;

[0045] S1-2, according to the Raman characteristic peaks of the cis isomer and the Raman characteristic peaks of the trans isomer, the Raman characteristic peak parameters of the plurality of mixtures are determined, and the Raman characteristic peak parameters are the Raman characteristic peak area of the cis isomer and the Raman characteristic peak area of the trans isomer;

[0046] For example, in combination with Figure 2 With Figure 3 As shown in S1-1~S1-2, the specific implementation process includes (1)~(3):

[0047] (1) Establish the standard Raman spectrum of cis-cyclopropane diethyl acid and trans-cyclopropane diethyl acid; take 1 mL of the standard sample of cis-cyclopropane diethyl acid or trans-cyclopropane diethyl acid on the surface of a clean silicon wafer, place the silicon wafer on the sample table of the Raman spectrometer, focus the 532 nm laser on the surface of the standard sample, use a laser power of 10 mW, set the spectral acquisition range to 150~2000 cm -1 , use 1200 lines / mm grating for spectral resolution optimization, and set the integration time to 30 s, perform 3 scans to obtain the standard Raman spectrum, and the results are shown in the standard Raman spectrum of cis and trans cyclopropane diethyl acid shown in Figure 2 ;

[0048] (2) Extract the Raman characteristic peaks of cis-cyclopropane diethyl acid and trans-cyclopropane diethyl acid; the obtained Raman spectrum of the cis and trans isomers is subjected to characteristic peak extraction; in combination with Figure 2 As shown in FIG. 2, 860 cm -1 , 980 cm -1 , and 1194 cm -1 are selected as the Raman characteristic peaks of cis-cyclopropane diethyl acid, and 252 cm -1 , 752 cm -1 , and 875 cm -1 are selected as the Raman characteristic peaks of trans-cyclopropane diethyl acid;

[0049] (3) Obtain the Raman spectrum and characteristic peak area of the isomer mixture with different mass fractions;

[0050] By preparing the standard sample of cis / trans cyclopropane diethyl acid mixture with a mass fraction of 10~90%, the Raman spectrum data is collected in the same way as step (1), and the characteristic peak area in step (2) is calculated by Gaussian fitting.

[0051] The range of the plurality of mass ratios is 1-9:10, and the Raman characteristic peak area of the cis isomer and the Raman characteristic peak area of the trans isomer are calculated by using the Gaussian fitting method in the art;

[0052] For example, the plurality of mass ratios are specifically implemented as follows: standard solutions with the total mass ratio of cis-diethyl cyclopropane diacid and trans-diethyl cyclopropane diacid being 10%, 20%, 30%, 40%, 50%, 60%, 70%, 80%, and 90% are prepared, and the Raman spectrum of each mass ratio sample is obtained by determining according to step (1) in the above content, as shown in FIG. 1. Figure 3 The Raman characteristic peaks of cis-diethyl cyclopropane diacid at 860 cm-1, 980 cm-1, and 1194 cm-1 and the Raman characteristic peaks of trans-diethyl cyclopropane diacid at 252 cm-1, 752 cm-1, and 875 cm-1 are calculated by Gaussian fitting, and the results are shown in Table 1. -1 -1 -1 -1 -1 -1

[0053] Table 1: Peak area calculation results under a plurality of mass ratios

[0054]

[0055] S2, based on the Raman characteristic peak parameters of a plurality of mixtures, a plurality of Raman characteristic peak ratio type variables between the cis isomer and the trans isomer are constructed; the Raman characteristic peak ratio type variables are used to describe the difference between the Raman characteristic peaks of the cis isomer and the trans isomer;

[0056] The Raman characteristic peak ratio type variables include original variables and constructed variables obtained by combining the original variables; and the original variables include single-peak ratio of Raman characteristic peak area, weighted combination ratio, logarithmic transformation, multi-peak product ratio, symmetry index, ratio of single-peak to weighted sum of multi-peak, combination construction of symmetric peak and asymmetric peak, cross ratio of homologous vibration modes in Raman spectrum, logarithmic or square root conversion of peak area;

[0057] The method for obtaining the constructed variables obtained by combining the original variables includes:

[0058] The original variables are fully combined in two-dimensional and three-dimensional combination spaces to generate a plurality of two-dimensional combinations and a plurality of three-dimensional combinations; the two-dimensional combinations and the three-dimensional combinations are used as the constructed variables; the specific operation steps include:

[0059] The number of variables in each dimension and the value set thereof are defined, and the values of each dimension are sequentially traversed to obtain all possible combinations: ​​​​​​

[0060] Two-dimensional output ordered pair: [(single peak ratio, weighted combination ratio), (weighted combination ratio, logarithmic transformation)];

[0061] Three-dimensional output triplet: [(single peak ratio, weighted combination ratio, logarithmic transformation), (single peak ratio, weighted combination ratio, multiplet product ratio)];

[0062] The result is a list containing all combinations, where each element is a tuple, which is a single-element tuple of the original variable itself for one-dimensional, a two-element tuple formed by the ordered pair of variables for two-dimensional, and a three-element tuple formed by the three variable elements for three-dimensional;

[0063] Specifically, the selection format of the Raman characteristic peak area ratio type variable is to form a characteristic matrix with the original variable or the constructed variable: where one-dimensional (the original variable itself) is a one-dimensional matrix; two-dimensional / three-dimensional form an n x kn x k matrix (k is 2 or 3); and the complete characteristic matrix and the target vector (the mass ratio of cis-cyclopropane diacid diethyl ester) are input into a linear regression model for global fitting:

[0064] Then, the R 2 value of the fitted curve is calculated and sorted, and at least one Raman characteristic peak area ratio type variable with the best fitting effect is screened out; for example, a multiple linear regression model is established by the least square method, the R 2 value of each combination is calculated, and the corresponding regression coefficient b is output, and multiple R 2 ≥ 0.99 model implementation results are shown in Table 2, where the best linear fitting ratio type variable is shown in Table 2. Figure 4

[0065] Table 2 Fitting results of ratio type variables

[0066]

[0067] S3, a prediction model is constructed, the input of the above prediction model is the above Raman characteristic peak ratio type variable, and the output is the mass ratio of the cis-isomer and the trans-isomer of cyclopropane diacid diethyl ester;

[0068] The method for constructing the prediction model comprises:

[0069] Correlation regression analysis is performed on the Raman characteristic peak area ratio type variable and multiple mass ratios, at least one ratio type variable with the highest correlation with the multiple mass ratios and the corresponding regression coefficient are selected, and a prediction model based on multiple linear regression is obtained;

[0070] The prediction model in the embodiment of the application adopts a multiple linear regression model; the function form is as shown in Table 2, which is a multiple linear regression model with the best sorting; Figure 4

[0071] ​​It should be understood that in other embodiments, any one of the following regression models can be selected: ordinary least squares (OLS), partial least squares regression (PLSR), principal component regression (PCR), ridge regression, Lasso regression, elastic net regression (Elastic Net), and support vector regression (SVR); the embodiments of the present application are not limited.

[0072] S4, based on the above prediction model, predicting the mass ratio of the cis-isomer and the trans-isomer of diethyl cyproheptadine in the diethyl cyproheptadine;

[0073] For example, a mixture of cis-isomer and trans-isomer of diethyl cyproheptadine to be detected is taken, and a sample of the mixture is tested according to step (1) in Embodiment 1, and the obtained Raman spectral characteristic peaks are analyzed, as shown in Table 1. Figure 5 The Raman characteristic peak area is calculated by Gaussian fitting, and R3=0.156, R6=1.138, and R7=0.047 are calculated, and then the prediction model Y=90.1294-82.2639R3-45.4856R6+280.6175R7 is used to calculate the mass fraction of cis-isomer as 38.7%.

[0074] Embodiment 2: The difference between this embodiment and Embodiment 1 is that the method for constructing the prediction model in S4 is different, which specifically includes:

[0075] Correlation analysis is performed on the Raman characteristic peak area proportional variable and the mass ratio of multiple cis-isomers and trans-isomers, and at least one Raman characteristic peak area proportional variable with the highest correlation with the mass ratio of multiple cis-isomers and trans-isomers is selected;

[0076] The at least one Raman characteristic peak area proportional variable with the highest correlation with the mass ratio of multiple cis-isomers and trans-isomers is used as the input of the prediction model, and the mass ratio of multiple cis-isomers and trans-isomers is used as the output, and a machine learning model is used for model training, and the model training method is the same as the existing training method. After training, the prediction model is constructed; the prediction model uses a random forest model; in other embodiments, one of a support vector regression model, a neural network model, etc. can be used.

[0077] In summary, the present application can realize the rapid and accurate detection of cis and trans isomer mixture without sample separation, at the same time, a variety of forms of combination variable is constructed and automatic traversal and screening in combination space is carried out, the optimal regression model is established to realize the high-precision quantitative prediction of isomer mass fraction, which has the advantages of non-destructive, rapid in-situ detection, compared with the traditional detection method, the Raman spectrum technology can directly analyze the sample without complex pretreatment steps, so as to realize the ultra-fast, non-destructive in-situ detection; It also has the high-precision quantitative modeling capability: the linear regression model is combined with the combination variable traversal screening strategy, the fitting ability of each group of characteristics and the cis mass fraction is evaluated, the prediction model is established, and the unknown sample can be quantitatively predicted with high sensitivity and high accuracy; Multi-component compatibility and strong universality: the method is not dependent on specific wavelength or specific structure type, can be applied to other structure similar molecular isomer system, has good universality and popularization value, and is suitable for configuration identification and quantification of complex systems such as drug synthesis intermediates and functional material precursors.

Claims

1. A method for the non-destructive detection of cis-trans- diethyl cyclopropane dicarboxylate based on a feature screening method, characterized in that, The method comprises the following steps: Obtaining Raman characteristic peak parameters of a plurality of mixtures formed by taking cis-isomer and trans-isomer of diethyl cyclopropanedicarboxylate in a plurality of mass ratios respectively; The method for obtaining Raman characteristic peak parameters of a plurality of mixtures comprises the following steps: Extracting Raman characteristic peaks of the cis-isomer from the standard Raman spectrum of the cis-isomer and extracting Raman characteristic peaks of the trans-isomer from the standard Raman spectrum of the trans-isomer; According to the Raman characteristic peaks of the cis-isomer and the trans-isomer, Raman characteristic peak parameters of a plurality of mixtures are determined, and the Raman characteristic peak parameters are Raman characteristic peak areas of the cis-isomer and the trans-isomer; Based on the Raman characteristic peak parameters of a plurality of mixtures, a plurality of Raman characteristic peak ratio type variables between the cis-isomer and the trans-isomer are constructed; the Raman characteristic peak ratio type variables are used to describe the difference between the Raman characteristic peaks of the cis-isomer and the trans-isomer; A prediction model is constructed, and the input of the prediction model is the Raman characteristic peak ratio type variables, and the output is the mass ratio of the cis-isomer and the trans-isomer of diethyl cyclopropanedicarboxylate; 2. A method for cis-trans detection of diethyl cyclohexenedioate based on feature screening according to claim 1, characterized in that, Based on the prediction model, the mass ratio of the cis-isomer and the trans-isomer of diethyl cyclopropanedicarboxylate is predicted.

3. A method for cis-trans detection of diethyl cyclohexenedioate based on feature screening as claimed in claim 1, wherein, The range of the plurality of mass ratios is 1-9:10, and the Raman characteristic peak areas of the cis-isomer and the trans-isomer are calculated by using a Gaussian fitting method.

4. A method for cis-trans detection of diethyl cyclohexane-1,2- dicarboxylate based on the characteristic screening according to claim 3, characterized in that, The Raman characteristic peak ratio type variables include original variables and constructed variables obtained by combining the original variables; the original variables include a unimodal ratio of Raman characteristic peak areas, a weighted combination ratio, a logarithmic transformation, a multi-peak product ratio, a symmetry index, a ratio of a weighted sum of unimodal and multi-peak, a combination construction of symmetric and asymmetric peaks, a cross ratio of homologous vibration modes in a Raman spectrum, a logarithmic or square root conversion of peak areas. The method for obtaining the constructed variables after the original variables are combined comprises the following steps:

5. A method for detecting cis-trans diethyl cyclohexane-1,2-dicarboxylate based on the characteristic screening according to claim 1, characterized in that, The original variables are subjected to full combination traversal in two-dimensional and three-dimensional combination spaces to generate a plurality of two-dimensional combinations and a plurality of three-dimensional combinations; and the plurality of two-dimensional combinations and the plurality of three-dimensional combinations are used as the constructed variables. The method for constructing the prediction model comprises the following steps: Correlation analysis is performed on the Raman characteristic peak area ratio type variables and a plurality of mass ratios of the cis-isomer and the trans-isomer, and at least one Raman characteristic peak area ratio type variable with the highest correlation with the plurality of mass ratios is selected; 6. A method for detecting cis-trans diethyl cyclohexane-1,2-dicarboxylate based on feature screening according to claim 1, characterized in that, The at least one Raman characteristic peak area ratio type variable with the highest correlation with the plurality of mass ratios is used as the input of the prediction model, the plurality of mass ratios of the cis-isomer and the trans-isomer are used as the output, a machine learning model is used for model training, and after the training is completed, the prediction model is constructed. The prediction model uses one of a support vector regression model, a random forest model and a neural network model.

Images