Method for determining moisture content of Xiao Chaihu granules based on near infrared spectroscopy

The near-infrared spectrum of Xiao Chaihu particles was corrected through the EPO algorithm, and a PLS model was established, which solved the problem of moisture content measurement under the influence of the differences in the origin of Chinese medicinal materials and the preparation process, and achieved rapid and accurate determination of the moisture content of Xiao Chaihu particles, improving the model stability and applicability.

CN115684062BActive Publication Date: 2024-11-19YUNNAN COMTESTOR CO LTD
View PDF 4 Cites 0 Cited by

Patent Information

Application Number
CN202211264887.6
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-10-17
Publication Date
2024-11-19
Estimated Expiration
2042-10-17

AI Technical Summary

Technical Problem

The existing method for determining moisture content of traditional Chinese medicine particles is complex and time-consuming, and cannot be applied to the rapid quality monitoring of small Chaihu particles. The applicability of the near-infrared spectral correction model is affected by the differences in the origin of Chinese medicine materials and the processing technology, resulting in stability and accuracy. Insufficient sex.

Method used

The near-infrared spectral data of small Chaihu particles were corrected by EPO algorithm, and the PLS model was established, and the spectral matrix of the 7500cm-1 to 4000cm-1 band was selected to build a stable and accurate moisture content correction model, and through multivariate scattering correction and Norris smoothing and second-order derivative preprocessing to reduce background information interference and improve the scope of application of the model.

Benefits of technology

The stability and accuracy of the model were significantly improved, with the determination coefficient R2 reaching 0.938, the RMSEC and RMSECV were 0.0429% and 0.0476%, respectively, and the average relative error of external verification was 1.92%, achieving a rapid and accurate determination of the moisture content of Xiao Chaihu particles. .

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN115684062B_ABST
    Figure CN115684062B_ABST
Patent Text Reader

Abstract

The present invention discloses a method for measuring the moisture content of Xiao Chaihu granules based on near infrared spectroscopy. The method is based on near infrared spectra of Xiao Chaihu granules of different batches, and uses EPO algorithm to correct 7500cm ‑1 ~4000 cm ‑1 The PLS model was established by using the band spectrum matrix to obtain the calibration model for determining the moisture content of Xiao Chaihu granules. The determination coefficient R of the model constructed by the present invention was 2 The optimal number of principal components is 8, the RMSEC and RMSECV are 0.0429% and 0.0476% respectively, the external validation prediction mean relative error PMRE is 1.92%, and there is no significant difference between the measured values ​​and the predicted values ​​of 60 external validation set samples at the significance level of α=0.05. The present invention provides an effective method for quickly and accurately determining the moisture content of Xiao Chaihu granules.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to a method for measuring the moisture content of Xiao Chaihu granules based on near infrared spectroscopy, belonging to the technical field of drug detection. Background Art

[0002] Xiao Chaihu Granules have the effects of dispelling heat, soothing the liver and stomach. They are widely used to treat symptoms such as alternating cold and heat, bitter fullness in the chest and flanks, loss of appetite, irritability and vomiting, bitter mouth and dry throat. They are made by boiling, extracting, concentrating, adding other auxiliary materials, granulating and drying. As a Chinese medicine granule preparation, the moisture content of Xiao Chaihu Granules is an important quality indicator. The Fischer method, drying method, vacuum drying method and toluene method are the methods for determining the moisture content of Chinese medicine granules included in the "Chinese Pharmacopoeia" (2020 edition). The above determination methods have the advantages of stable and accurate measurement, but they also have the disadvantages of complex pretreatment and time-consuming determination, which cannot be applied to the rapid quality monitoring of Xiao Chaihu Granules.

[0003] In recent years, with the improvement of the processing level of Chinese medicine preparations, process quality monitoring and finished product quality inspection have become important ways to further improve the quality of Chinese medicine preparations. As a green, rapid, non-destructive and online analysis method that can realize qualitative and quantitative detection at the same time, near-infrared spectroscopy analysis has been widely used in quality monitoring during the processing of Chinese medicine preparations. There are also relevant literature reports on the use of near-infrared spectroscopy to determine the moisture content of Chinese medicine preparations during processing and finished products. The accuracy and stability of the near-infrared correction model determine the promotion and application value of this technology. The accuracy and stability of the correction model are jointly determined by factors such as instrument firmware, spectral acquisition method, basic data determination method, modeling and model evaluation method, and sample size. The scope of application of the model is determined by the representativeness of the calibration set or training set samples. Due to the wide distribution of Chinese medicinal materials and differences in processing technology, the differences in near-infrared spectral information of different batches of Chinese medicinal materials will affect the applicability of the near-infrared correction model of Chinese medicine preparations.

[0004] When using near-infrared spectral information to establish a moisture content correction model for Xiao Chaihu Granules, how to effectively reduce the interference of "background" irrelevant information such as different origins of Chinese medicinal materials and differences in processing techniques on the correction model, and directly extract absorbance information related to moisture content can improve the stability, accuracy and scope of application of the model. Summary of the invention

[0005] The present invention provides a method for determining the moisture content of Xiao Chaihu granules based on near infrared spectroscopy with good stability and high accuracy. The method is based on the near infrared spectra of Xiao Chaihu granules of different batches of raw materials, and the EPO algorithm is used to correct the 7500cm -1 ~4000 cm -1The PLS model was established based on the band spectrum matrix to obtain the calibration model for determining the moisture content of Xiao Chaihu Granules, providing an effective method for rapid and accurate determination of the moisture content of Xiao Chaihu Granules.

[0006] The method for measuring the moisture content of Xiao Chaihu granules by EPO-PLS based on near infrared spectroscopy of the present invention is as follows:

[0007] 1. Collect Xiao Chaihu granules with a moisture content ranging from 1.8% to 2.8%, and measure the moisture content of Xiao Chaihu granules to obtain moisture content data;

[0008] 2. Use a near-infrared spectrometer to collect the near-infrared spectrum of Xiao Chaihu granules with a water content in the range of 1.8-2.8%;

[0009] 3. Preprocess the near-infrared spectrum collected in step 2 by multivariate scattering correction, Norris smoothing, and second-order derivative, and select the wavelength of the preprocessed near-infrared spectrum to be between 4000 and 7500 cm -1 The spectral data were randomly divided into calibration set spectral data and external validation set spectral data;

[0010] 4. Use the EPO algorithm to calibrate the calibration set spectral data and combine it with its corresponding moisture content data to obtain the calibration set data;

[0011] 5. Based on the calibration data, the partial least squares method was used to construct the moisture content of Xiao Chaihu granules (4000-7500cm -1 Near infrared correction model for the band;

[0012] 6. Import the external validation set spectral data corrected by the EPO algorithm into the Xiao Chaihu granules moisture 4000-7500cm -1 The near-infrared correction model of the band is used to obtain its moisture content; for the Xiao Chaihu granules to be tested, its near-infrared spectrum is collected, and after spectral preprocessing and band selection according to the method in step 3, the data is imported into the Xiao Chaihu granules moisture 4000-7500cm -1 The near-infrared correction model of the band can be used to obtain its water content.

[0013] The EPO algorithm is used to correct the spectral data:

[0014] 1. Decompose the matrix X composed of the near-infrared spectra of m Xiao Chaihu granules with different moisture contents in the range of 1.8%-2.8% into a moisture content correlation matrix X1, a moisture content uncorrelated matrix X2, and a residual matrix X3, wherein X1=XO, X2=XQ, the dimension of the matrix X is m×n, n is the number of absorbances of the wavelength points of the near-infrared spectrum, O is the projection matrix of the moisture content correlation matrix, and Q is the projection matrix of the moisture content uncorrelated matrix;

[0015] X=X1+X2+X3 (1)

[0016] X=XO+XQ+X3 (2);

[0017] 2. Calculate the difference matrix D of the near-infrared spectra of m Chaihu granules with different moisture contents in the range of 1.8%-2.8%, D=X i -X0, where X i is the near infrared spectrum of the i-th moisture content sample, and X0 is the near infrared spectrum of the first moisture content sample;

[0018] 3. Perform principal component analysis on the difference matrix D in step 2, that is, D=TP T Decompose the difference matrix D, where T is the score matrix and P is the loading matrix;

[0019] 4. Select the principal component number C whose extraction rate of matrix data by principal component analysis is greater than 99%, and calculate the load matrix P of principal component number C. C ;

[0020] 5. Through P c P c T Calculate the moisture content related projection matrix of X1;

[0021] 6. Use the m×n unit matrix to subtract the projection matrix of the X1 moisture content correlation matrix in step (5) to obtain a new moisture content correlation projection matrix, C=IQ, where I is the m×n unit matrix;

[0022] 7. Transform the original matrix based on the new moisture content related projection matrix obtained in step 6, that is, X1=XC.

[0023] Advantages and technical effects of the present invention:

[0024] 1. The EPO correction used in the method of the present invention can significantly improve the stability and accuracy of the model. EPO + wavelength (7500cm -1 ~4000cm -1 ) has the best modeling effect, and the determination coefficient of the model is R 2 The optimal number of principal components is 8, RMSEC and RMSECV are 0.0429% and 0.0476% respectively, the external validation prediction mean relative error PMRE is 1.92%, and there is no significant difference between the measured and predicted values ​​of the 60 external validation set samples at the significance level of α=0.05;

[0025] 2. The method of the present invention improves the accuracy and stability of the near-infrared spectroscopy method for determining the moisture content of multiple batches of Xiao Chaihu granules. The method of the present invention is simple, efficient, and has a value for promotion and application. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 The statistical analysis results of the moisture content of 420 Xiao Chaihu granules;

[0027] Figure 2 Near infrared spectra of original (a) and treated (b) Xiao Chaihu granules;

[0028] Figure 3 RMSECV under different c and number of principal components;

[0029] Figure 4 The scores of the first and second principal components of the principal component analysis of the calibration set sample spectrum after preprocessing;

[0030] Figure 5 The first and second principal component scores of the principal component analysis of the calibration set sample spectrum after EPO correction;

[0031] Figure 6 The characteristic spectrum of water (4980-5140cm -1 and 6900-7010cm -1 )’s first and second principal component scores of principal component analysis;

[0032] Figure 7 The scores of the first and second principal components of the principal component analysis of the EPO-corrected water characteristic spectrum;

[0033] Figure 8 This is the root mean square error RMSECV result diagram of the model interaction validation under different principal component numbers;

[0034] Fig. 9 Uncorrected 7500-4000 cm -1 PLS model constructed by near infrared spectrum of the band;

[0035] Fig.10 7500~4000 cm after EPO correction -1 PLS model constructed by near infrared spectrum of the band;

[0036] Fig.11 PLS model constructed for the uncorrected near-infrared spectra of the characteristic absorption band of water;

[0037] Fig.12 PLS model constructed for the near-infrared spectrum of the characteristic absorption band of water after EPO correction. DETAILED DESCRIPTION

[0038] The present invention will be further described below in conjunction with the accompanying drawings and specific embodiments, but the protection scope of the present invention is not limited thereto. The methods in the embodiments are conventional methods unless otherwise specified.

[0039] Instruments used in the examples: AUW220 analytical balance (Shimadzu, Japan); FED400 multifunctional hot air circulation oven (BINDER, Germany); near infrared spectrometer: Nicolet Antaris II including an integrating sphere diffuse reflectance analysis module with an InGaAs detector, a rotator, a 5 cm quartz cup and other sampling accessories and RESULT TM Spectral acquisition software (ThermoFisher Scientific, USA);

[0040] Example 1: 70 batches of Xiaohu Granules (National Medicine Standard Z53021588) produced by Yunnan Baiyao Co., Ltd. from April to December 2021 were collected, with a sampling volume of 6 per batch, for a total of 420 samples;

[0041] 1. Determination of moisture content of Xiao Chaihu granules

[0042] The moisture content of the Xiao Chaihu granules was determined by referring to the second method (drying method) of 0832 Moisture Determination Method in Part IV of the General Rules of the Chinese Pharmacopoeia (2020 Edition); the results are as follows Figure 1 As shown by Figure 1 It can be seen that the average moisture content of 420 Xiao Chaihu granules is 2.23%, the maximum and minimum values ​​are 1.49% and 3.34% respectively, and the moisture content of most Xiao Chaihu granules is between 1.70% and 2.76%.

[0043] 2. Collection of near infrared spectra

[0044] A Nicolet Antaris II near-infrared spectrometer was used, and the main parameters of the instrument were spectral scanning range 10000~4000 cm -1 ; Resolution: 8 cm -1 ; Scan times: 64 times. After 3 hours of preheating, the near-infrared spectra of the samples were collected. Three spectra were collected for each sample, and the average spectrum was taken as the modeling spectrum of the sample. The results are shown in Figure 2 a;

[0045] 3. Spectral preprocessing and selection of modeling bands

[0046] TQ Analyst 9.3 (Thermo Fisher Scientific, USA) was used to preprocess the near-infrared spectra collected in step 2 by multiplicative scattering correction, Norris smoothing, and second-order derivative. Multiplicative Scatter Correction (MSC) can eliminate the scattering effect caused by the uneven distribution of sample particles; Norris smoothing and second-order derivative (SD) can filter noise and eliminate baseline drift. The Norris smoothing window width was 11 and the window interval was 3. The wavelength of the preprocessed near-infrared spectrum was selected to be between 4000 and 7500 cm -1 The spectral data of 420 samples were divided into 360 calibration set samples and 60 external validation samples by using Kennard-Stone algorithm.

[0047] The near infrared spectrum after preprocessing is shown in Figure 2 b, by Figure 2 b It can be seen that the absorbance difference of different samples at the absorption peak position is larger than that of the original spectrum after the sample spectrum is preprocessed by MSC+Norris+SD, and the 9000-1000cm -1 There is high frequency noise, so 4000-7500cm is selected -1 The band is the full wavelength range modeled in this embodiment;

[0048] At the same time, the characteristic absorption peak of moisture at 5154 cm in the near-infrared spectrum of Xiao Chaihu granules was -1 and 6944 cm -1 4980-5140cm -1 and 6900-7010cm -1 The characteristic spectral band of moisture in Xiao Chaihu granules was used as a control;

[0049] 4. Use the EPO algorithm to calibrate the calibration set spectral data, and combine it with the corresponding moisture content data to obtain the calibration set data, and use the "rchemo" software package in the R4.1 software to perform spectral EPO correction;

[0050] The near infrared spectrum matrix of samples with water contents of 1.7%, 1.8%, 1.9%, 2.0%, 2.1%, 2.2%, 2.3%, 2.4%, 2.5%, 2.6%, and 2.69% was used as the EPO analysis matrix; the near infrared spectrum matrix X (m×n), m=11;

[0051] The EPO algorithm decomposes the matrix X (m×n) composed of the near-infrared spectra of the 11 moisture content Xiao Chaihu granules into: moisture content correlation matrix X1, moisture content irrelevant matrix X2, and residual matrix X3, where X1=XO, X2=XQ, matrix X is m×n, n is the number of absorbances of the near-infrared spectrum wavelength points, O is the projection matrix of the moisture content correlation matrix, and Q is the projection matrix of the moisture content irrelevant matrix;

[0052] X=X1+X2+X3;

[0053] X=XO+XQ+X3;

[0054] The model's "leave one out" cross-validation root mean square error RMSECV minimum value is used to optimize the optimal number of EPO factors C and the optimal number of principal components of the PLS regression model. The results are shown in Figure 3 , EPO calibration full band (4000-7500cm -1 ) are 1, and the optimal number of main factors for the PLS regression model is 6;

[0055] The purpose of the EPO algorithm is to solve the moisture content correlation matrix P. The steps for solving the matrix P include:

[0056] (1) Calculate the difference matrix D of the near-infrared spectra of m Chaihu granules with different moisture contents in the range of 1.8%-2.8%, D = X i -X0, where X i is the near infrared spectrum of the i-th moisture content sample, and X0 is the near infrared spectrum of the first moisture content sample;

[0057] (2) Perform principal component analysis on the difference matrix D in step (1), that is, D = TP T Decompose the difference matrix D, where T is the score matrix and P is the loading matrix;

[0058] (3) Select the principal component number C whose extraction rate of matrix data by principal component analysis is greater than 99%, and calculate the loading matrix P of principal component number C C ;

[0059] (4) Through P c P c T Calculate the moisture content related projection matrix of X1;

[0060] (5) Use the m×n unit matrix to subtract the projection matrix of the X1 moisture content correlation matrix in step (5) to obtain a new moisture content correlation projection matrix, C=IQ, where I is the m×n unit matrix;

[0061] (6) Transform the original matrix based on the moisture content related projection matrix obtained in step (5), that is, X1=XC;

[0062] Results Figure 4 , 5 , 6, 7, by Figure 5-7 It can be seen that compared with the uncorrected spectrum, the uniformity of the spatial distribution of the first and second principal component scores of the spectrum matrix after EPO correction is improved. The uniform distribution of the sample principal component scores is beneficial to improving the stability of the correction model.

[0063] 5. Establishment and evaluation of the moisture content correction model of Xiao Chaihu granules before and after EPO correction

[0064] The sampling partial least squares (PLS) method was used to establish the near-infrared correction model of the full band and the characteristic absorption band of water before and after EPO correction;

[0065] Evaluation of the accuracy and stability of the near-infrared calibration model: (1) Model parameters: determination coefficient (R 2 ), Root Mean Standard Error of Calibration (RMSEC) and Root Mean Square Error of Cross Validation (RMSECV); (2) Mean relative error of prediction (PMRE) of the external validation set and the paired t-test results between the predicted values ​​and the measured values; the above models were established and evaluated using the libPLS package of Matlab software;

[0066] Model interaction validation is used to optimize the optimal number of PLS ​​principal components. The RMSECV of the four models changes with the number of principal components. Figure 8 shown by Figure 8 It can be seen that the optimal number of principal components of the full-band PLS model before and after EPO correction is 8, and the optimal number of principal components of the moisture characteristic spectrum before and after EPO correction is 11 and 9 respectively;

[0067] The model parameters and the validation results of 60 external validation samples are shown in Tables 1 and 2, respectively. Figures 9 to 12 As shown;

[0068] Table 1 Parameters of the near-infrared correction model for moisture content of Xiao Chaihu granules established under different modeling input variables

[0069]

[0070] As shown in Table 1, EPO correction improves the stability and accuracy of the full-band and moisture characteristic spectral correction models to a certain extent. The stability and accuracy of the full-band PLS model are better than those of the moisture characteristic band PLS model. The stability and accuracy of the PLS model established by EPO correction for the full band are the best among the four models, with a determination coefficient of 0.938, an optimal number of principal factors of 8, and mean square estimation residuals RMSEC and interactive validation mean square error RMSECV of 0.0429 and 0.0476, respectively.

[0071] Table 2 External validation results of the near-infrared correction model for moisture content of Xiao Chaihu granules established under different modeling input variables

[0072]

[0073] As shown in Table 2, the external validation results of the four models show that the minimum PMRE of the full-band model corrected by EPO is 1.92%, and the maximum of the water characteristic spectrum is 3.24%. The paired t-test results of the predicted values ​​and measured values ​​of 60 samples show that there is no significant difference between the two values ​​at the α=0.05 level. Based on the model parameters and external validation results, the near-infrared PLS correction model of the moisture content of Xiao Chaihu granules with the full-band model corrected by EPO is the best.

[0074] In summary, MSC+Norris(11,3)+SD were used to preprocess the near-infrared spectrum of Xiao Chaihu Granules, and the PLS model was established after the original spectrum matrix was corrected by EPO based on the near-infrared spectrum matrix of Xiao Chaihu Granules with different moisture contents. The results showed that EPO correction can significantly improve the stability and accuracy of the near-infrared correction model of Xiao Chaihu Granules moisture. This study provides a useful technical reference for the rapid and accurate determination of the moisture content of Xiao Chaihu Granules.

Claims

1. A method for measuring the moisture content of Xiao Chaihu granules based on near infrared spectroscopy, characterized in that: Here are the steps: (1) Collecting Xiao Chaihu granules with a moisture content ranging from 1.8% to 2.8%, and measuring the moisture content of the Xiao Chaihu granules to obtain moisture content data; (2) A near-infrared spectrometer was used to collect the near-infrared spectra of Xiao Chaihu granules with a water content ranging from 1.8% to 2.8%. (3) Preprocess the near-infrared spectrum collected in step (2) by multivariate scattering correction, Norris smoothing, and second-order derivative, and select the wavelength of the preprocessed near-infrared spectrum to be between 4000 and 7500 cm -1 The spectral data were randomly divided into calibration set spectral data and external validation set spectral data; (4) Using the EPO algorithm to calibrate the calibration set spectral data, and combining it with its corresponding moisture content data to obtain the calibration set data; (5) Based on the calibration data, the partial least squares method was used to construct the moisture content of Xiao Chaihu granules (4000-7500 cm -1 Near infrared correction model for the band; (6) Import the spectral data of the external validation set corrected by the EPO algorithm into the moisture content of Xiao Chaihu granules 4000-7500cm -1 The near infrared correction model of the band is used to obtain its moisture content; for the Xiao Chaihu granules to be tested, its near infrared spectrum is collected, and after spectral preprocessing and band selection according to the method in step (3), the data is imported into the Xiao Chaihu granules moisture 4000-7500cm -1 The near-infrared correction model of the band is used to obtain its water content; The EPO algorithm is used to correct the spectral data: (1) The matrix X consisting of the near-infrared spectra of m Xiao Chaihu granules with different moisture contents in the range of 1.8%-2.8% is decomposed into a moisture content correlation matrix X1, a moisture content irrelevant matrix X2, and a residual matrix X3, where X1=XO, X2=XQ, the dimension of the matrix X is m×n, n is the number of absorbances at the wavelength points of the near-infrared spectrum, O is the projection matrix of the moisture content correlation matrix, and Q is the projection matrix of the moisture content irrelevant matrix; X=X1+X2+X3; X=XO+XQ+X3; (2) Calculate the difference matrix D of the near-infrared spectra of m Chaihu granules with different moisture contents in the range of 1.8%-2.8%, D = X i -X0, where X i is the near infrared spectrum of the i-th moisture content sample, and X0 is the near infrared spectrum of the first moisture content sample; (3) Perform principal component analysis on the difference matrix D in step (2), that is, D = TP T Decompose the difference matrix D, where T is the score matrix and P is the loading matrix; (4) Select the principal component number C whose extraction rate of matrix data by principal component analysis is greater than 99%, and calculate the loading matrix P of principal component number C C ; (5) Through P c P c T Calculate the moisture content related projection matrix of X1; (6) Use the m×n unit matrix to subtract the projection matrix of the X1 moisture content correlation matrix in step (5) to obtain a new moisture content correlation projection matrix, C=IQ, where I is the m×n unit matrix; (7) Based on the moisture content related projection matrix obtained in step (6), the original matrix is ​​transformed, that is, X1 = XC.

Citation Information

Patent Citations

  • Method for measuring content of effective constituent of common goldenrop granules

    CN110320286A

  • Method for predicting basic density of papermaking wood chips through near infrared spectrum for eliminating moisture interference

    CN114813463A

  • Method for correcting measuring-point-free temperature compensation model during online application of near infrared spectrum analyzer

    WO2017084118A1

  • Qualitative and quantitative combined method for constructing near infrared quantitative model

    WO2018010352A1