CN113376118A - Scutellaria baicalensis near-infrared online quality detection method based on partial least squares regression method - Google Patents

Scutellaria baicalensis near-infrared online quality detection method based on partial least squares regression method Download PDF

Info

Publication number
CN113376118A
CN113376118A CN202110221861.2A CN202110221861A CN113376118A CN 113376118 A CN113376118 A CN 113376118A CN 202110221861 A CN202110221861 A CN 202110221861A CN 113376118 A CN113376118 A CN 113376118A
Authority
CN
China
Prior art keywords
scutellaria baicalensis
sample
wave band
infrared
value
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202110221861.2A
Other languages
Chinese (zh)
Inventor
蔡宝昌
刘晓
王天舒
金俊杰
秦昆明
李伟东
杨超
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Nanjing Haichang Chinese Medicine Group Co ltd
Nanjing Haiyuan Chinese Herbal Pieces Co ltd
Original Assignee
Nanjing Haichang Chinese Medicine Group Co ltd
Nanjing Haiyuan Chinese Herbal Pieces Co ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Nanjing Haichang Chinese Medicine Group Co ltd, Nanjing Haiyuan Chinese Herbal Pieces Co ltd filed Critical Nanjing Haichang Chinese Medicine Group Co ltd
Priority to CN202110221861.2A priority Critical patent/CN113376118A/en
Publication of CN113376118A publication Critical patent/CN113376118A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
    • G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
    • G01N21/25—Colour; Spectral properties, i.e. comparison of effect of material on the light at two or more different wavelengths or wavelength bands
    • G01N21/31—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry
    • G01N21/35—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry using infrared light
    • G01N21/359—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry using infrared light using near infrared light
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N1/00—Sampling; Preparing specimens for investigation
    • G01N1/28—Preparing specimens for investigation including physical details of (bio-)chemical methods covered elsewhere, e.g. G01N33/50, C12Q

Landscapes

  • Physics & Mathematics (AREA)
  • Spectroscopy & Molecular Physics (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Chemical & Material Sciences (AREA)
  • Analytical Chemistry (AREA)
  • Biochemistry (AREA)
  • General Health & Medical Sciences (AREA)
  • General Physics & Mathematics (AREA)
  • Immunology (AREA)
  • Pathology (AREA)
  • Investigating Or Analysing Materials By Optical Means (AREA)

Abstract

The invention discloses a scutellaria baicalensis near-infrared online quality detection method based on a partial least squares regression method, which comprises the following steps of: (1) sample preparation: taking scutellaria baicalensis decoction piece samples of different producing areas and different batches; (2) collecting near infrared spectrum data: simultaneously collecting and recording a near-infrared spectrogram of a scutellaria baicalensis sample and a near-infrared spectrogram of scutellaria baicalensis after powdering; (3) preprocessing the spectral data: respectively adopting an original spectrum, a first order derivation, a second order derivation, a multivariate scattering correction, a vector normalization, a convolution smoothing filter, a multivariate scattering correction and a vector normalization, a convolution smoothing filter and a multivariate scattering correction, and a convolution smoothing filter and a vector normalization to preprocess the near infrared spectrum data of the scutellaria baicalensis sample before and after powdering; (4) and establishing a scutellaria baicalensis quantitative correction model by adopting a partial least squares regression method. The established model is accurate and reliable, the operation is quick and simple, and the content of baicalin and water in the scutellaria baicalensis decoction pieces can be directly measured without damage.

Description

Scutellaria baicalensis near-infrared online quality detection method based on partial least squares regression method
Technical Field
The invention belongs to the technical field of medicinal material detection; in particular to a Partial Least Squares Regression (PLSR) -based method for detecting the near-infrared quality of scutellaria baicalensis.
Background
The Chinese medicinal material Scutellariae radix is Scutellaria baicalensis Georgi of LabiataeScutellaria baicalensisGeorgi's dried root, bitter in taste, cold in nature, enters the lung, gallbladder, spleen, small intestine and large intestine channels, and has the effects of clearing heat and drying dampness, purging fire and removing toxicity, stopping bleeding and preventing miscarriage. Scutellaria baicalensis has a long history of medication, is recorded in Shen nong Ben Cao Jing for the earliest time, and is one of the most commonly used traditional Chinese medicines in the clinical practice of traditional Chinese medicine. Therefore, the perfection of the quality detection method of the scutellaria has positive significance on the aspects of clinical curative effect of the traditional Chinese medicine, stable quality of the Chinese patent medicine and the like.
At present, the quality control of scutellaria baicalensis is mainly characterized by character identification, microscopic identification, TLC qualitative research, extract determination and a content determination method based on HPLC. The TLC method is a main detection method for qualitatively identifying scutellaria baicalensis, and adopts a polyamide film, ethyl acetate-methanol is used as a solvent for extraction, and toluene-ethyl acetate-methanol-formic acid is selected as a development system. The content determination research of the active ingredient baicalin is more. Besides the liquid phase fingerprint, researchers establish the medicinal material fingerprint of the scutellaria baicalensis by various methods such as liquid chromatography-mass spectrometry, capillary electrophoresis chromatography, ultraviolet spectrum, infrared spectrum, nuclear magnetic resonance and the like, and comprehensively control the quality of the medicinal material of the scutellaria baicalensis.
The invention adopts a near-infrared quality detection method based on a partial least squares regression method (PLSR), establishes an extract production online detection system through links such as raw material quality detection, extraction process detection and the like, can effectively solve the defects of inconvenient sampling, low efficiency, environmental pollution and the like in production detection, and improves the product quality; meanwhile, the method provides guidance for reaction termination, reduces energy consumption and realizes green production of the extract.
Disclosure of Invention
In order to solve the technical problems, the invention provides a Partial Least Squares Regression (PLSR) -based method for detecting the near-infrared quality of scutellaria baicalensis. The quality detection method established by the invention is quick and simple to operate, and the established model is accurate and reliable and can be used for quantitative analysis of the content of baicalin and water in the scutellaria baicalensis decoction pieces.
In order to achieve the above purpose, the technical scheme provided by the invention is as follows:
a scutellaria baicalensis near-infrared online quality detection method based on a partial least squares regression method is characterized by comprising the following steps:
(1) sample preparation: taking scutellaria baicalensis decoction piece samples of different producing areas and different batches;
(2) collecting near infrared spectrum data: simultaneously collecting and recording a near-infrared spectrogram of a scutellaria baicalensis sample and a near-infrared spectrogram of scutellaria baicalensis after powdering;
(3) preprocessing the spectral data: respectively adopting an original spectrum, a first order derivation, a second order derivation, a multivariate scattering correction, a vector normalization, a convolution smoothing filter, a multivariate scattering correction and a vector normalization, a convolution smoothing filter and a multivariate scattering correction, and a convolution smoothing filter and a vector normalization to preprocess the near infrared spectrum data of the scutellaria baicalensis sample before and after powdering;
(4) and establishing a scutellaria baicalensis quantitative correction model by adopting a partial least squares regression method.
As a preferred scheme, in the above method for online quality detection of scutellaria baicalensis near infrared based on partial least squares regression, the method for acquiring near infrared spectrum in step (2) is as follows: adding about 10 g of the pulverized sample into a quartz sample tube, and filling and flattening the sample; selecting a flat sample from the unfired sample, and enabling the flat sample to be fully connected with the near-infrared diffuse reflection optical fiber probeAnd (4) contacting. The test environment temperature is 25 ℃, and the relative humidity is 45-60%; taking the built-in background of the instrument as a reference, deducting the background, and collecting in an integrating sphere diffuse reflection mode in a wave number range of 12000-4000 cm-1Resolution of 8 cm-1The number of scanning times is 64, each sample is scanned for 2 times, and the average spectrum is taken as the near infrared spectrum of the sample.
Optimized screening of Partial Least Squares Regression (PLSR) model parameters
The quantitative model design of the near infrared spectrum adopts Python programming language, the integrated development environment is Pycharm consistency, and the operating system is Windows 10.
Before a quantitative correction model is established, the original spectrum needs to be preprocessed, so that the influence of many factors such as high-frequency noise, scattered light, stray light, a sample state, instrument response and the like in the measuring process can be avoided. The spectrum preprocessing can remove unnecessary information and improve the prediction accuracy of the model. The model of the invention screens the optimal spectrum pretreatment method aiming at different components before and after powdering, which comprises the following steps: raw spectra (Spectrum), First derivative (1 stD), Second derivative (2 stD), Multivariate Scattering Correction (MSC), vector normalization (SNV), convolution smoothing filter (Savitzky-Golay filter, S-G), multivariate scattering correction + vector normalization, convolution smoothing filter + multivariate scattering correction, convolution smoothing filter + vector normalization.
In addition, the invention selects proper spectrum wave band, which can reduce redundant information in spectrum and improve the prediction precision of model. Meanwhile, when the PLSR method is used for modeling, different principal component numbers have great influence on the model prediction result. If the number of principal components is too high, an "overfitting" phenomenon occurs, but if the number of principal components is too small, the spectrum information used is too small. The R value, the Root Mean Square Error (RMSE) and the corrected mean square error (RMSEC) are used as indexes, the optimal pretreatment method of the powder-baicalin is vector normalization, and the optimal spectral band is 6206.760-5053.360 cm-1 The selected number of the main components is 5; the optimal pretreatment method of powder-water content comprises convolution smoothing filtering and multi-element scattering correction, and the optimal spectrum wave bandIs 5049.502-4281.855 cm-1 The number of selected main components is 1; the optimal pretreatment method of the decoction piece-baicalin is first-order derivation, and the optimal spectral band is 7364.018-6982.123 cm-1 The number of selected principal components is 3; the optimal pretreatment method of the decoction pieces and the water comprises multivariate scattering correction and vector normalization, and the optimal spectral band is 8714.153-8525.134 cm-1 The number of selected principal components is 5. Preferred results of the calibration model and the evaluation parameters are shown in table 1.
TABLE 1 PLSR model and evaluation parameters
Figure RE-GDA0003135973710000031
Has the advantages that: compared with the prior art, the invention has the advantages that:
the invention adopts Fourier transform Near Infrared (NIR) analysis technology to collect Scutellariae radix decoction pieces and near infrared spectrogram after powdering, screens optimal pretreatment method of different component original spectra before and after powdering Scutellariae radix, optimizes main factor number and selects optimal wave band, and establishes quantitative analysis model by Partial Least Squares Regression (PLSR) method. The verification result shows that the model established by the invention has high accuracy, quick and simple operation, does not need an extraction process, can nondestructively judge the quality of the scutellaria baicalensis, and can accurately detect the baicalin and the water content in the scutellaria baicalensis.
Drawings
FIG. 1 is a near infrared spectrum of pulverized Scutellariae radix decoction pieces;
FIG. 2 is a near infrared spectrum of Scutellariae radix decoction pieces without pulverizing;
FIG. 3 shows HPLC chromatograms (1. baicalin) of Scutellariae radix sample (A) and control (B).
Detailed Description
The present invention will be described below with reference to specific examples to make the technical aspects of the present invention easier to understand and grasp, but the present invention is not limited thereto. The experimental methods described in the following examples are all conventional methods unless otherwise specified; the reagents and materials are commercially available, unless otherwise specified.
Example 1
1. Experimental Material
1.1 test drugs
The total amount of Scutellariae radix decoction pieces is 100 batches, and all the decoction pieces are from Nanjing sea-sourced traditional Chinese medicine decoction piece Co., Ltd, as shown in Table 1.
TABLE 1100 sources of Baical skullcap root decoction pieces
Producing area Gansu Hebei river Heilongjiang (Jilin) Inner cover Shanxi province
Number of 10 23 8 14 19 26
1.1 laboratory instruments and reagents
Bruker-sensor 37 fourier transform mid-ir and near-ir spectrometer, including OPUS5.0 software, Pbs detector (Bruker, germany); waters e2695 high performance liquid chromatograph (Waters corporation, USA) Waters2998 ultraviolet detector; one in ten thousand balance BSA2245-CW (Beijing Saedodus scientific instruments, Inc.); a one-hundred-thousandth balance model AG-285 (METTLER TOLEDO, Switzerland); KY-500E ultrasonic cleaner (Kunshan ultrasonic Instrument Co., Ltd.); HH-6 digital display constant temperature water bath (Changzhou national electric appliance Co., Ltd.); Milli-Q ultra pure water instruments (Millipore, USA); GeneSpeed X1 microcentrifuge (International trade for genetic Biotechnology, Shanghai, Inc.).
Baicalin reference (batch No. P20A9F59353, content ≥ 98%) was purchased from Shanghai-sourced leaf Biotech limited. Phosphoric acid was chromatographically pure (Shanghai Aladdin science and technology Co., Ltd.), methanol was chromatographically pure (TEDIA Co., USA), and anhydrous ethanol was analytically pure (Susheng chemical Co., Ltd., Wuxi city).
2. Experimental methods and results
2.1 acquisition of the near Infrared Spectrum
Pulverizing 100 batches of Scutellariae radix decoction pieces, sieving with No. 5 sieve, and measuring the near infrared spectra of pulverized 100 batches of Scutellariae radix decoction pieces and pulverized Scutellariae radix decoction pieces. Adding about 10 g of the pulverized sample into a quartz sample tube, and filling and flattening the sample; and selecting a flat sample from the samples which are not pulverized, so that the sample can be fully contacted with the near-infrared diffuse reflection optical fiber probe. The test environment temperature is 25 ℃, and the relative humidity is 45-60%. And taking the background in the instrument as a reference, and subtracting the background. The collection mode is diffuse reflection of an integrating sphere, and the wave number range is 12000-4000 cm-1Resolution of 8 cm-1The number of scanning times is 64, each sample is scanned for 2 times, and the average spectrum is taken as the near infrared spectrum of the sample. The near infrared spectra of pulverized and un-pulverized Scutellariae radix decoction pieces are shown in FIG. 1 and FIG. 2, respectively.
2.2 determination of the content of baicalin
2.2.1 preparation of test solutions
The preparation of the test solution is carried out according to the requirements of the Baikal skullcap root content determination item of the first part of China pharmacopoeia 2020 edition:
taking about 0.3g of the powder in the product, accurately weighing, adding 40ml of 70% ethanol, heating and refluxing for 3 hours, cooling, filtering, placing the filtrate in a 100ml measuring flask, washing the container and the residue with a small amount of 70% ethanol in several times, filtering the washing solution in the same measuring flask, adding 70% ethanol to the scale, and shaking up. Precisely measuring 1ml, placing into a 10ml measuring flask, adding methanol to scale, and shaking.
2.2.2 preparation of control solution A proper amount of baicalin control dried at 60 deg.C under reduced pressure for 4 hr was precisely weighed and added with methanol to make into solution containing 60 μ g per 1 ml.
2.2.3 chromatographic conditions using octadecylsilane chemically bonded silica as filler; methanol-water-phosphoric acid (47: 53: 0.2) is used as a mobile phase; the detection wavelength was 280 nm. The number of theoretical plates is not less than 2500 calculated by baicalin peak. The amount of the sample was 10. mu.L. The chromatogram is shown in FIG. 3.
2.2.4 preparation of Standard Curve
Sampling 6 standard solutions with different concentration levels, i.e. 0.065375mg/ml, 0.0326875 mg/ml, 0.0163435 mg/ml, 0.0081715 mg/ml, 0.0040855 mg/ml and 0.0020425 mg/ml, respectively, detecting, drawing a standard curve by taking a peak area (Y) as an ordinate and baicalin concentration (X, mg/ml) as an abscissa, and obtaining a regression equation of the baicalin, namely Y =3.12 × 107X+6.90×103R is 0.9997, and the linear range is 20.43-653.75 mg/mL.
2.3 moisture determination of Scutellariae radix decoction pieces
The water content of 100 batches of scutellaria baicalensis decoction pieces is measured according to a second drying method of a four-part water content measuring method (general rule 0832) in 2020 edition of Chinese pharmacopoeia.
3. Establishment of scutellaria baicalensis near infrared spectrum quantitative model
3.1 modeling result after adding SPXY Algorithm
PLSR: and (3) determining the number of main components and an optimal preprocessing method (a correction set and a verification set are obtained by adopting a spxy algorithm, wherein the proportion of the correction set is 80 percent, and the proportion of the verification set is 20 percent). And (5) solving R2 values, R2 values, RMSEP values and RMSEC values of the prediction set under different main component numbers and different preprocessing methods. Prediction set R2The higher the value, the lower the RMSEP value the better the model.
3.1.1 modeling of Scutellaria baicalensis Georgi powdering and baicalin content
The parameter is the main component of 6, the SNV + SG is preprocessed, and the whole spectrum is obtained. 80% training, 20% testing. The error of 20 samples of the set was tested. The mean absolute error is 0.0000885, and the mean absolute value of the relative error is 0.0529. Prediction set R2Value, correction set R2The values, RMSEP values and RMSEC values were 0.4747, 0.5329, 0.0090 and 0.0213, respectively.
3.1.2 Scutellaria baicalensis powdering and moisture modeling
The parameter is the main component of 4, the MSC is preprocessed, and the whole spectrum is taken. 80% training, 20% testing. The error of 20 samples of the set was tested. The mean absolute error was-0.0006 and the mean absolute value of the relative error was 0.0410. The prediction set R2 values, correction set R2 values, RMSEP values, and RMSEC values were 0.3638, 0.3398, 0.0036, and 0.0077, respectively.
3.1.3 modeling of the content of the undried baical skullcap root and the baicalin
The parameter is the principal component of 2, the pretreatment is the first order difference, and the whole spectrum is taken. 80% training, 20% testing. The error of 20 samples of the set was tested. The average absolute error was-0.0011 and the average of the absolute values of the relative errors was 0.0889. The prediction set R2 values, correction set R2 values, RMSEP values, and RMSEC values were 0.4607, 0.4734, 0.0133, and 0.0188, respectively.
3.1.4 modeling of radix Scutellariae without powdering and moisture
The parameter is taken as the main component 2, the main component is preprocessed into MSC + SNV, and the whole spectrum is taken. 80% training, 20% testing. The error of 20 samples of the set was tested. The mean absolute error was 0.0005 and the mean absolute value of the relative error was 0.0445. The prediction set R2 values, correction set R2 values, RMSEP values, and RMSEC values were 0.0063, 0.0540, 0.0039, and 0.0094, respectively.
3.2 modeling results after adding SPXY segmentation Algorithm
PLSR: and (3) determining an optimal wave band, an optimal main component number and an optimal preprocessing method (a correction set and a verification set are obtained by adopting a spxy algorithm, the proportion of the correction set is 80%, and the proportion of the verification set is 20%). Comparing the prediction set R under different wave bands, different main component numbers and different preprocessing methods2The value is obtained. The closer to 1, the better the result.
3.2.1 powdering-content
The modeling result when the wave band length is 50 is that when the wave band is 1600-1649, the number of the principal components is 2, and the preprocessing method is SNV, the result is optimal, and the prediction set R is2The value is 0.5294.
The modeling result when the wave band length is 100-2The value is 0.2635.
When the wave band length is 150, the modeling result is optimal when the wave band is 300-2The value is 0.3079.
The modeling result when the wave band length is 200, when the wave band is 1600-When the method is SNV, the result is optimal, and the set R is predicted2The value is 0.4601.
The modeling result when the wave band length is 250 is that when the wave band is 1500-1749, the number of the principal components is 7, and the preprocessing method is the original spectrum, the result is optimal, and the prediction set R is2The value is 0.3900.
The modeling result when the wave band length is 300 is that when the wave band is 1500-2The value is 0.6521.
The modeling result when the wave band length is 350 is that when the wave band is 1400-1749, the number of the principal components is 5, and the preprocessing method is MSC, the result is optimal, and the prediction set R is2The value is 0.4156.
The modeling result when the wave band length is 400 is that when the wave band is 1600-2The value is 0.5306.
The modeling result when the wave band length is 450, when the wave band is 1350-2The value is 0.5822.
The modeling result when the wave band length is 500 is optimal when the wave band is 1500-2The value is 0.4617.
The modeling result when the wave band length is 550 is that when the wave band is 1100-1649, the number of the principal components is 11, and the preprocessing method is the original spectrum, the result is optimal, and the prediction set R is2The value is 0.3986.
The modeling result when the wave band length is 600 shows that when the wave band is 1200-1799, the number of the principal components is 9, and the preprocessing method is the original spectrum, the result is optimal, and the prediction set R is2The value is 0.4899.
In conclusion, when the length of the wave band is 300, the screening 1500-1799 wave band is optimal, the number of the main components is 5, and when the pretreatment method is SNV, 80% of training and 20% of testing are performed. The error of 20 samples in the test set was-0.0018 on average and 0.0430 on average of the absolute value of the relative error. Prediction set R2Value, correction set R2The values, RMSEP and RMSEC were 0.6521, 0.7365, 0.0073 and 0.0159, respectively.
3.2.2 powdering-moisture
The modeling result when the wave band length is 50, when the wave band is 1800-1849, the number of the main components is 4, and the preprocessing method is MSC + convolution smoothing, the result is optimal, and the prediction set R is2The value was 0.3926.
According to the modeling result when the waveband length is 100, when the waveband is 1800-1899, the number of the principal components is 3, and the preprocessing method is SNV + convolution smoothing, the result is optimal, and the value of the prediction set R2 is 0.1784.
When the wave band is 31800-2The value is 0.2497.
The modeling result when the wave band length is 200 is the optimal result when the wave band is 1800 + 1999, the number of the main components is 1 and the preprocessing method is MSC + convolution smoothing, and the prediction set R2The value is 0.4922.
The modeling result when the wave band length is 250 is the optimal result when the wave band is 1750-2The value is 0.4331.
The modeling result when the wave band length is 300, when the wave band is 1500-2The value is 0.2163.
When the wave band length is 1750-2074, the number of the principal components is 6, and the preprocessing method is convolution smoothing, the result is optimal, and the prediction set R is2The value is 0.3312.
The modeling result when the wave band length is 400 is that when the wave band is 1600-2The value is 0.4754.
The modeling result when the wave band length is 450, when the wave band is 1350-2The value is 0.1844.
The modeling result when the wave band length is 500 is that when the wave band is 1500-2The value is 0.4036.
The modeling result when the wave band length is 550 is that when the wave band is 1650-2074, the number of the main components is 1, and the preprocessing method is MSC + convolution smoothing, the result is optimal, and the prediction set R is2The value was 0.4318.
The modeling result when the wave band length is 600, when the wave band is 1800-2074, the number of the principal components is 2, and the preprocessing method is convolution smoothing, the result is optimal, and the prediction set R is2The value was 0.0847.
The modeling result when the wave band length is 650, when the wave band is 1300-2The value is 0.4460.
The modeling result when the wave band length is 700 is that when the wave band is 1400-2074, the number of the principal components is 1, and the preprocessing method is SNV + convolution smoothing, the result is optimal, and the prediction set R is optimal2The value is 0.1461.
The modeling result when the wave band length is 750, when the wave band is 1500-containing 2074, the number of the main components is 1, and the preprocessing method is MSC, the result is optimal, and the prediction set R is2The value is 0.4005.
The modeling result when the wave band length is 800 is that when the wave band is 1600-2074, the number of the principal components is 1, and the preprocessing method is MSC + convolution smoothing, the result is optimal, and the prediction set R is2The value is 0.4704.
In conclusion, when the band length is 150, the 1800 + 1949 band is selected to be optimal, the number of main components is 1, and when the preprocessing method is MSC + convolution smoothing, 80% of training and 20% of testing are performed. The error of 20 samples in the test set was 0.0003 in average absolute error and 0.0353 in average absolute value of relative error. Prediction set R2Value, correction set R2The values, RMSEP and RMSEC were 0.4922, 0.1340, 0.0032 and 0.0087, respectively.
3.2.3 Unpowdering-content
The modeling result when the wave band length is 50 shows that when the wave band is 1250-At the time of initial spectrum, the result is optimal, and a prediction set R2The value is 0.5905.
When the wave band is 1200-1299, the number of the main components is 3, and the preprocessing method is the first-order difference, the result is optimal, and the prediction set R is obtained according to the modeling result when the wave band length is 1002The value is 0.6091.
The modeling result when the wave band length is 150 shows that when the wave band is 1650-1799, the number of the main components is 4, and the preprocessing method is SNV, the result is optimal, and the prediction set R is2The value is 0.5580.
The modeling result when the wave band length is 200 is that when the wave band is 1600-2The value was 0.5517.
The modeling result when the wave band length is 250 is that when the wave band is 1000-charge 1249, the number of the principal components is 6, and the preprocessing method is the original spectrum, the result is optimal, and the prediction set R is2The value is 0.4748.
The modeling result when the wave band length is 300, when the wave band is 1500-2The value is 0.5451.
The modeling result when the wave band length is 350 is that when the wave band is 1750-2The value is 0.5188.
The modeling result when the wave band length is 400 shows that when the wave band is 1200-1499, the number of the principal components is 11, and the preprocessing method is the original spectrum, the result is optimal, and the prediction set R is2The value is 0.4668.
When the wave band is 1350-2The value is 0.4898.
When the wave band is 1500-1999, the number of the main components is 5, the preprocessing method is the first-order difference, the result is optimal, and the prediction set R is2The value is 0.4292.
The modeling result when the band length is 550 is that when the band is 1100-1649 and the number of principal components is 2, the number of the principal components is pre-determinedWhen the processing method is a first-order difference, the result is optimal, and the set R is predicted2The value is 0.4710.
The modeling result when the wave band length is 600 shows that when the wave band is 1200-1799, the number of the principal components is 5, and the preprocessing method is convolution smoothing, the result is optimal, and the prediction set R is2The value is 0.4241.
In conclusion, when the length of the band is 100, the 1200-1299 band is selected to be optimal. The number of main components is 3, and when the pretreatment method is first-order difference, 80% of training and 20% of testing are carried out. The error of 20 samples in the test set was-0.0025 on average, and 0.0713 on average. Prediction set R2Value, correction set R2The values, RMSEP and RMSEC were 0.6091, 0.2447, 0.0113 and 0.0227, respectively.
3.2.4 Unpowdering-moisture
And (3) as a modeling result when the wave band length is 50, when the wave band is 850-899, the number of the main components is 5, and the preprocessing method is MSC + SNV, the result is optimal, and the value of the prediction set R2 is 0.2463.
According to the modeling result when the wave band length is 100, when the wave band is 800-899, the number of the principal components is 4, and the preprocessing method is SNV + convolution smoothing, the result is optimal, and the value of the prediction set R2 is 0.2425.
According to the modeling result when the wave band length is 150, when the wave band is 750-899, the principal component number is 6, and the preprocessing method is SNV, the result is optimal, and the value of the prediction set R2 is 0.2361.
According to the modeling result when the wave band length is 200, when the wave band is 1200-1399, the number of the main components is 3, and the preprocessing method is convolution smoothing, the result is optimal, and the value of the prediction set R2 is 0.1781.
And (3) as a modeling result when the wave band length is 250, when the wave band is 1500-1749, the number of the principal components is 5, and the preprocessing method is convolution smoothing, the result is optimal, and the value of the prediction set R2 is 0.1432.
According to the modeling result when the wave band length is 300, when the wave band is 1200-1499, the number of the principal components is 3, and the preprocessing method is convolution smoothing, the result is optimal, and the value of the prediction set R2 is 0.1670.
And (3) as a modeling result when the waveband length is 350, when the waveband is 1400-1749, the number of the principal components is 7, and the preprocessing method is convolution smoothing, the result is optimal, and the value of the prediction set R2 is 0.1294.
According to the modeling result when the wave band length is 400, when the wave band is 1200-1599, the number of the principal components is 3, and the preprocessing method is convolution smoothing, the result is optimal, and the value of the prediction set R2 is 0.1652.
And (3) as a modeling result when the wave band length is 450, when the wave band is 1350-.
According to the modeling result when the wave band length is 500, when the wave band is 1500-.
According to the modeling result when the wave band length is 550, when the wave band is 1650-2074, the number of the main components is 1, and the preprocessing method is SNV, the result is optimal, and the value of the prediction set R2 is 0.0512.
According to the modeling result when the waveband length is 600, when the waveband is 1800-2074, the number of the main components is 1, and the preprocessing method is SNV, the result is optimal, and the value of the prediction set R2 is 0.0436.
In conclusion, when the length of the band is 50, the 850-899 band is selected to be optimal. The number of main components is 5, and when the pretreatment method is MSC + SNV, 80% of training and 20% of testing are performed. The error of 20 samples of the test set was-0.00009420 on average and 0.0368 on average of the absolute value of the relative error. The prediction set R2 values, correction set R2 values, RMSEP values, and RMSEC values were 0.2463, 0.0922, 0.0034, and 0.0094, respectively.
Test set samples not participating in the modeling were externally validated. And inputting the sample into a quantitative model to obtain a predicted value, and inspecting the prediction capability of the model through the relative deviation of the predicted value and a true value obtained by a conventional method. The test results are shown in table 2. The average value of the absolute value of the relative error between the predicted value and the true value of the water model established by the scutellaria baicalensis powder is 3.53 percent; the average value of the absolute value of the relative error between the predicted value and the true value of the water model established by the scutellaria baicalensis decoction pieces is 3.68 percent; the average value of the relative error absolute value of the predicted value and the true value of the baicalin content model established by the baical skullcap root powder is 4.3 percent; the average value of the relative error absolute value of the predicted value and the true value of the baicalin model established by the baical skullcap root decoction pieces is 7.13%. The results show that the relative error between the predicted value and the true value of the moisture is small, the prediction result is accurate, and the model is successfully established. The relative error between the predicted value and the true value of the baicalin content is small, and the prediction result is accurate.
TABLE 2 test set sample prediction results
Figure DEST_PATH_IMAGE004
The above detailed description is specific to one possible embodiment of the present invention, and the embodiment is not intended to limit the scope of the present invention, and all equivalent implementations or modifications without departing from the scope of the present invention should be included in the technical scope of the present invention.

Claims (3)

1. A scutellaria baicalensis near-infrared online quality detection method based on a partial least squares regression method is characterized by comprising the following steps:
(1) sample preparation: taking scutellaria baicalensis decoction piece samples of different producing areas and different batches;
(2) collecting near infrared spectrum data: simultaneously collecting and recording a near-infrared spectrogram of a scutellaria baicalensis sample and a near-infrared spectrogram of scutellaria baicalensis after powdering;
(3) preprocessing the spectral data: respectively adopting an original spectrum, a first order derivation, a second order derivation, a multivariate scattering correction, a vector normalization, a convolution smoothing filter, a multivariate scattering correction and a vector normalization, a convolution smoothing filter and a multivariate scattering correction, and a convolution smoothing filter and a vector normalization to preprocess the near infrared spectrum data of the scutellaria baicalensis sample before and after powdering;
(4) and establishing a scutellaria baicalensis quantitative correction model by adopting a partial least squares regression method.
2. The scutellaria baicalensis near-infrared online quality detection method based on the partial least squares regression method according to claim 1, characterized in that the method for collecting near-infrared spectra in the step (2) is as follows: adding about 10 g of the pulverized sample into a quartz sample tube, and filling and flattening the sample; selecting a flat sample from the samples without being powdered, enabling the sample to be in full contact with the near-infrared diffuse reflection optical fiber probe,
the test environment temperature is 25 ℃, and the relative humidity is 45-60%; taking the built-in background of the instrument as a reference, deducting the background, and collecting in an integrating sphere diffuse reflection mode in a wave number range of 12000-4000 cm-1Resolution of 8 cm-1The number of scanning times is 64, each sample is scanned for 2 times, and the average spectrum is taken as the near infrared spectrum of the sample.
3. The method for detecting the quality of the scutellaria baicalensis near infrared online based on the partial least squares regression method as claimed in claim 1, wherein the partial least squares regression method establishes the optimal parameters of the scutellaria baicalensis quantitative correction model as follows:
PLSR model and evaluation parameters
Figure 613323DEST_PATH_IMAGE002
。
CN202110221861.2A 2021-02-27 2021-02-27 Scutellaria baicalensis near-infrared online quality detection method based on partial least squares regression method Pending CN113376118A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110221861.2A CN113376118A (en) 2021-02-27 2021-02-27 Scutellaria baicalensis near-infrared online quality detection method based on partial least squares regression method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110221861.2A CN113376118A (en) 2021-02-27 2021-02-27 Scutellaria baicalensis near-infrared online quality detection method based on partial least squares regression method

Publications (1)

Publication Number Publication Date
CN113376118A true CN113376118A (en) 2021-09-10

Family

ID=77569634

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110221861.2A Pending CN113376118A (en) 2021-02-27 2021-02-27 Scutellaria baicalensis near-infrared online quality detection method based on partial least squares regression method

Country Status (1)

Country Link
CN (1) CN113376118A (en)

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
李化等: "基于近红外漫反射光谱和多元数据分析的黄芩质量标准的快速评价方法研究", 《药物分析杂志》 *

Similar Documents

Publication Publication Date Title
CN102288572A (en) Method for quickly detecting content of index ingredient of traditional Chinese medicinal material by utilizing near infrared spectrum technique
CN113960241A (en) Method for rapidly screening illegal addition of prohibited chemicals in health care products based on thin-layer chromatography-near infrared spectrum coupling technology
CN111272941B (en) Ultra-high performance liquid chromatography determination method for bear gall powder in phlegm-heat clear injection
CN111007190B (en) Method for constructing UPLC (ultra performance liquid chromatography) characteristic spectrum of rhizoma bolbostemmae medicinal material and method for measuring component content of rhizoma bolbostemmae medicinal material
CN113376116A (en) Near-infrared online quality detection method for rehmannia
CN109030410B (en) Construction method of royal jelly near-infrared quantitative correction model and royal jelly detection method
CN116429943A (en) Quality analysis method of traditional Chinese medicine radix scutellariae based on dual-wavelength equal-baseline differential fusion map
CN111855867B (en) Method for establishing characteristic spectrum of traditional Chinese medicine or traditional Chinese medicine composition preparation and application
CN107271396A (en) The quick determination method of general flavone content in a kind of tealeaves
CN113376118A (en) Scutellaria baicalensis near-infrared online quality detection method based on partial least squares regression method
CN113376117A (en) Near-infrared online quality detection method for angelica sinensis
Long et al. A simple and effective method for identification of Fraxini Cortex from different sources by multi‐mode fingerprint combined with chemometrics
CN110031577B (en) Quality detection method and identification application of traditional Chinese medicine or traditional Chinese medicine composition preparation
CN110672751B (en) UPLC characteristic spectrum establishing method and detecting method for fresh houttuynia cordata medicinal material
CN110320286B (en) Method for measuring content of active ingredients of Xiaochaihu granules
CN112834456A (en) Near-infrared online quality detection method for white paeony roots
CN108535213B (en) Method for evaluating potential use of effective components of cassia twig and monitoring production process
CN112113929A (en) Quality control method for extraction process of children's oral liquid for removing food retention and relieving cough
Zhu et al. Rapid Quality Identification of Decoction Pieces of Crude and Processed Corydalis Rhizoma by Near-Infrared Spectroscopy Coupled with Chemometrics
CN109030408A (en) A kind of quick catechu medicinal material detection method
CN113376119A (en) Near-infrared online quality detection method for bighead atractylodes rhizome
CN116106467B (en) Method for identifying raw pinellia tuber in Huoxiang Zhengqi water
CN109342357A (en) The construction method of Zhenqi Fuzheng prepn near-infrared quantitative calibration models and the detection method of Zhenqi Fuzheng prepn
CN113655136B (en) Characteristic spectrum of lung-heat clearing and toxin expelling granule and construction method thereof
CN115856150B (en) Fingerprint construction method and identification method of traditional Chinese medicine and traditional Chinese medicine preparation

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
RJ01 Rejection of invention patent application after publication
RJ01 Rejection of invention patent application after publication

Application publication date: 20210910