Method for rapidly characterizing physical parameters of traditional Chinese medicine powder

Abstract

The invention provides a method for rapidly characterizing physical parameters of traditional Chinese medicine powder. The method comprises the following steps: (1) determining a near infrared spectrum of a traditional Chinese medicine powder sample; (2) measuring physical parameters of the traditional Chinese medicine powder sample, and constructing a physical fingerprint spectrum; and (3) preprocessing the near infrared spectrum, screening spectrum variables related to the physical parameters, establishing a spectrum variable model, and representing the physical parameters. According to the method for rapidly representing the physical parameters of the traditional Chinese medicine powder, rapid conversion of the physical fingerprint spectrum of the powder is achieved; after model performance investigation, the GRNN model is found to have the best prediction capability on the physical parameters of the medicinal material powder, and the GRNN simplified model still shows satisfactory prediction performance; the result proves that the NIR spectrum technology has great application potential in the rapid characterization of the physical properties of the pellet core and the traditional Chinese medicine powder.

Description

Technical Field

[0001] The invention belongs to the field of traditional Chinese medicine, and in particular relates to a method for quickly characterizing physical parameters of traditional Chinese medicine powder. Background Art

[0002] In the production process of traditional Chinese medicine preparations, powdering medicinal materials is generally a necessary step before subsequent processing. Due to the complexity of traditional Chinese medicine itself, the quality of different batches of medicinal materials often fluctuates. In the relevant research on medicinal material quality control, chromatographic fingerprint based on chemical composition information is one of the most widely used technologies. However, there are very few research reports related to the physical properties of medicinal material powders. However, it should be emphasized that when the batch-to-batch differences in the physical characteristics of medicinal material powders are too large, it often affects the subsequent production process and thus affects the quality of the final product. Therefore, it is necessary to use appropriate technologies to accurately characterize the physical properties of medicinal material powders, obtain prior knowledge of the physical information of medicinal materials in advance, and effectively regulate the physical parameters of the samples. Summary of the invention

[0003] In view of this, the present invention aims to overcome the defects in the prior art and propose a method for rapidly characterizing the physical properties of traditional Chinese medicine powder.

[0004] To achieve the above object, the technical solution of the present invention is achieved as follows:

[0005] A method for rapidly characterizing physical parameters of traditional Chinese medicine powders comprises the following steps:

[0006] (1) Determine the near infrared spectrum of traditional Chinese medicine powder samples;

[0007] (2) determining the physical parameters of the Chinese medicine powder sample and constructing a physical fingerprint;

[0008] (3) Preprocessing the near-infrared spectrum, screening spectral variables related to the physical parameters, establishing a spectral variable model, and characterizing the physical parameters.

[0009] Furthermore, the Chinese medicine powder sample in step (1) is at least one of Caulis Sinomenii, Scutellaria baicalensis, Pueraria lobata or Paeonia lactiflora.

[0010] Furthermore, the particle size of the traditional Chinese medicine powder sample is 0.1-355 μm.

[0011] Furthermore, the near infrared spectrum acquisition condition in step (1) is Empty 1×-16cm -1 Resolution - 32 scans, Empty 2× - 8cm -1 Resolution - 64 scans, Empty 2× - 16cm -1Resolution - 64 scans or Empty 2× - 8cm -1 Resolution - At least one of the 32 scans.

[0012] Preferably, the near infrared spectrum acquisition condition in step (1) is Empty 2×-8cm -1 Resolution - 64 scans.

[0013] Furthermore, the physical parameter in step (2) is at least one of bulk density, tap density, angle of repose, angle of collapse, loss on drying, hygroscopicity, porosity between powders, Carr index, Hausner ratio, percentage of particles less than 50 μm, or particle size uniformity.

[0014] Furthermore, the preprocessing method in step (3) is at least one of SNV, SG smoothing, 1st Der-SG smoothing or SNV-1st Der-SG smoothing.

[0015] Furthermore, the method for establishing the spectral variable model in step (3) is at least one of DS, PLSR or GRNN.

[0016] Preferably, the method for establishing the spectral variable model in step (3) is PLSR or GRNN.

[0017] Compared with the prior art, the present invention has the following advantages:

[0018] The method for rapidly characterizing the physical parameters of Chinese medicine powders described in the present invention realizes the rapid conversion of the physical fingerprint of the powders; after investigating the model performance, it is found that the GRNN model has the best prediction ability for the physical parameters of medicinal material powders, and the simplified GRNN model still shows satisfactory prediction performance; the results confirm that NIR spectroscopy technology has great application potential in the rapid characterization of the physical properties of pill cores and Chinese medicine powders. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 The RSD curves of the spectrum of medicinal powder under different spectrum collection conditions described in Example 1 of the present invention are as follows: Figure A is white peony root; Figure B is scutellaria baicalensis; Figure C is Sinomenium sibiricum; Figure D is Pueraria root;

[0020] Figure 2 The Euclidean distance of the spectrum of the medicinal material powder under different spectrum collection conditions described in Example 1 of the present invention: wherein, Figure A is white peony root; Figure B is scutellaria baicalensis; Figure C is Sinomenium sibiricum; Figure D is Pueraria root;

[0021] Figure 3The physical fingerprints of the four medicinal material training set samples described in Example 2 of the present invention are as follows: A is white peony root; B is scutellaria baicalensis; C is Caulis Sinomenii; and D is pueraria root (Note: α: angle of repose; c: angle of collapse; Da: bulk density; Dc: tap density; IC: Carr index; IH: Hausner ratio; Ie: porosity between powders; %HR: loss on drying; %H: hygroscopicity; %pf: percentage of particle size <50 μm; Iθ: particle size uniformity);

[0022] Figure 4 The correlation coefficients between the physical parameters of the four medicinal material powders described in Example 2 of the present invention are shown in Figure A: White Peony Root; Figure B: Scutellaria baicalensis; Figure C: Caulis Sinomenii; Figure D: Pueraria lobata;

[0023] Figure 5 The NIR pre-processed spectra of the four medicinal powders described in Example 3 of the present invention are shown in Figure A: White Peony Root; Figure B: Scutellaria baicalensis; Figure C: Caulis Sinomenii; Figure D: Pueraria lobata;

[0024] Figure 6 The prediction accuracy of the physical parameters of the white peony validation set samples based on different models described in Example 3 of the present invention: wherein, Figure A is the angle of repose; Figure B is the collapse angle; Figure C is the bulk density; Figure D is the tap density; Figure E is the Carr index; Figure F is the Hausner ratio; Figure G is the porosity between powders; Figure H is the drying loss; Figure I is the hygroscopicity; Figure J is the percentage of particle size less than 50 μm; Figure K is the particle size uniformity (Note: DS: DS model based on original spectral variables; VS-DS: DS model based on key spectral variables; PLS: PLS regression model based on original spectral variables; VS-PLS: PLS regression model based on key spectral variables; GRNN: GRNN model based on original spectral variables; VS-GRNN: GRNN model based on key spectral variables);

[0025] Figure 7 The prediction accuracy of the physical parameters of the Scutellaria baicalensis validation set samples based on different models described in Example 3 of the present invention is as follows: wherein, Figure A is the angle of repose; Figure B is the collapse angle; Figure C is the bulk density; Figure D is the tap density; Figure E is the Carr index; Figure F is the Hausner ratio; Figure G is the porosity between powders; Figure H is the drying loss; Figure I is the hygroscopicity; Figure J is the percentage of particle size less than 50 μm; Figure K is the particle size uniformity;

[0026] Figure 8 The prediction accuracy of physical parameters of the Caulis Sinensis validation set samples based on different models described in Example 3 of the present invention is as follows: wherein, Figure A is the angle of repose; Figure B is the angle of collapse; Figure C is the bulk density; Figure D is the tap density; Figure E is the Carr index; Figure F is the Hausner ratio; Figure G is the porosity between powders; Figure H is the loss on drying; Figure I is the hygroscopicity; Figure J is the percentage of particles less than 50 μm; Figure K is the particle size uniformity;

[0027] Fig. 9 The prediction accuracy of physical parameters of the Pueraria root validation set samples based on different models described in Example 3 of the present invention is as follows: wherein, Figure A is the angle of repose; Figure B is the collapse angle; Figure C is the bulk density; Figure D is the tap density; Figure E is the Carr index; Figure F is the Hausner ratio; Figure G is the porosity between powders; Figure H is the drying loss; Figure I is the hygroscopicity; Figure J is the percentage of particle size less than 50 μm; Figure K is the particle size uniformity;

[0028] Fig.10 Predicted values ​​and actual values ​​of physical parameters of medicinal material powder based on the GRNN model described in Example 3 of the invention: wherein, Figure A is white peony root; Figure B is scutellaria baicalensis; Figure C is Sinomenium scutellariae; Figure D is Pueraria root;

[0029] Fig.11 The physical fingerprint prediction map and actual map of the Chinese medicine powder verification set samples based on the GRNN model described in Example 3 are shown: wherein, Figure A1 is the physical fingerprint prediction map of the white peony powder sample; Figure A2 is the physical fingerprint actual map of the white peony powder sample; Figure B1 is the physical fingerprint prediction map of the scutellaria powder sample; Figure B2 is the physical fingerprint actual map of the scutellaria powder sample; Figure C1 is the physical fingerprint prediction map of the Sinomenium sinomenium powder sample; Figure C2 is the physical fingerprint actual map of the Sinomenium sinomenium powder sample; Figure D1 is the physical fingerprint prediction map of the kudzu root powder sample; Figure D2 is the physical fingerprint actual map of the kudzu root powder sample;

[0030] Fig.12 The near-infrared absorption peak attribution analysis results of the key variables of the medicinal materials described in Example 3 are shown: Figure A is white peony root; Figure B is scutellaria baicalensis; Figure C is Sinomenium scutellariae; and Figure D is Pueraria lobata. DETAILED DESCRIPTION

[0031] Unless otherwise defined, the technical terms used in the following examples have the same meanings as those generally understood by those skilled in the art to which the present invention belongs. The test reagents used in the following examples, unless otherwise specified, are all conventional biochemical reagents; the experimental methods, unless otherwise specified, are all conventional methods.

[0032] Materials and experimental equipment:

[0033] Caulis Sinomenium, Radix Scutellariae, Radix Puerariae and Radix Paeoniae Alba were purchased from the Bozhou Medicinal Materials Market, and 13 batches of each were purchased. Caulis Sinomenium was identified by Hu Yunfei, a teacher at Bozhou University, as the dried vine stem of Sinomenium acutum (Thunb.) Rehd. et Wils., a plant of the Menispermaceae family; Radix Scutellariae was identified as the dried root of Scutellaria baicalensis Georgi, a plant of the Lamiaceae family; Radix Puerariae was identified as the dried root of Pueraria lobata (Willd.) Ohwi, a plant of the Leguminosae family; and Radix Paeoniae Alba was identified as the dried root of Paeonia lactiflora Pall., a plant of the Ranunculaceae family.

[0034] Antaris II Fourier transform near-infrared spectrometer (Thermo Fisher Scientific, USA); BT-2600 laser particle size distribution analyzer (Dandong Better Instrument Co., Ltd., China); BT-1001 intelligent powder property tester (Dandong Better Instrument Co., Ltd., China); 150SD pharmaceutical stability test chamber (Beijing Lanbeishi Constant Temperature Technology Co., Ltd., China); FW135 Chinese herbal medicine pulverizer (Tianjin Test Instrument Co., Ltd., China); JJ1000 electronic balance (Changshu Shuangjie Testing Instrument Factory, China); AL204 electronic balance (Mettler-Toledo International AG, Switzerland); HX204 moisture analyzer (Mettler-Toledo International AG, Switzerland).

[0035] Sample preparation:

[0036] Each batch of medicinal materials was crushed for 12 seconds and then sieved using a standard drug sieve (in compliance with the 2020 edition of the Pharmacopoeia) into three samples with particle sizes ranging from 250-355 μm, 150-250 μm, and less than 150 μm. Finally, 39 batches of powder samples were prepared for each medicinal material, for a total of 156 batches.

[0037] The present invention will be described in detail below with reference to the embodiments.

[0038] Example 1 Near infrared spectrum collection

[0039] 2 g of each medicinal material was randomly and precisely weighed from three samples of different particle size ranges, and then mixed to obtain four mixed samples. A full factorial experimental design with three factors and three levels was conducted with the number of spectral scans, resolution, and gain as the investigation factors, and a total of 27 spectral acquisition conditions were obtained, as shown in Table 1. Six spectra of the mixed samples were repeatedly collected using an NIR spectrometer, and the spectral acquisition conditions of each medicinal material were optimized using the RSD value and Euclidean distance of the spectrum as evaluation indicators. Then the NIR spectrum of the sample was collected.

[0040] Table 1 Factors and levels considered in the optimization of NIR spectral acquisition conditions

[0041]

[0042] Under each spectral acquisition condition, the RSD value of the absorbance of the six NIR spectra of the mixed sample at each wavelength was calculated. The results are as follows Figure 1 As shown. It should be noted that when optimizing the spectral acquisition conditions of Scutellaria baicalensis, Pueraria lobata and Sinomenium sinensis samples, the spectral information could not be acquired when the gain was set to Empty 4×. Therefore, for these three herbs, only 18 spectral acquisition conditions were investigated. The smaller the RSD, the better the stability of the spectrum. Figure 1 The results show that the best spectral acquisition conditions for the powders of four medicinal materials, namely, Sinomenium scabra, White Peony Root, Scutellaria baicalensis, and Pueraria lobata, are “Empty 1×-16cm -1 Resolution - 32 scans", "Empty 2× - 8cm -1 Resolution - 64 scans", "Empty 2× - 16cm -1 Resolution - 64 scans", "Empty 2× - 8cm -1 Resolution - 32 scans". At the same time, the Euclidean distances of the six NIR spectra under each condition were calculated ( Figure 2 ). The results showed that the spectral acquisition conditions of Sinomenium sinomenium, Paeonia lactiflora and Puerariae lobata were consistent with the analysis results based on the RSD curve, which indicated that the three optimized spectral acquisition conditions were scientific and reliable. -1 When the condition is “Empty 2×-8cm”, the Euclidean distance of the spectrum of the Scutellaria baicalensis sample is the smallest. -1 There was no significant difference in the spectral Euclidean distance when the resolution was 64 scans, so this study selected the latter as the optimal spectral acquisition condition for Scutellaria baicalensis samples. After the optimization of the NIR spectral acquisition conditions, spectra of 156 samples were collected.

[0043] Example 2 Construction of physical fingerprint and analysis of correlation between physical parameters

[0044] The parameters for constructing the physical fingerprint of medicinal powder include bulk density, tap density, angle of repose, angle of collapse, loss on drying, hygroscopicity, porosity between powders, Carr index, Hausner ratio, percentage of particles less than 50 μm, and particle size uniformity, a total of 11 parameters. The method for determining the percentage of sample particles less than 50 μm and the particle size uniformity parameter is as follows: add an appropriate amount of sample to the storage funnel of the laser particle size distribution analyzer, calculate the percentage of powder with a particle size less than 50 μm after sampling at a constant speed, and then read the particle size uniformity parameter of the sample. The calculation formula of the sample physical parameters and the numerical conversion formula are shown in Table 2. After determining the physical parameters of all samples, correlation analysis is performed on them.

[0045] Table 2 Parameters and calculation formulas used to construct physical fingerprints of Chinese medicine powders

[0046]

[0047] The measured physical parameters of the samples were processed according to the conversion formula in Table 2, and then the physical fingerprint was prepared. The physical fingerprints of the training set samples of the four medicinal materials are shown in Figure 2. Figure 3 As shown in the figure. As can be seen from the figure, the angle of repose, collapse angle and Hausner ratio in the physical fingerprint of white peony sample are larger than those of other medicinal materials, which indicates that the fluidity of white peony powder is better. This is because the texture of white peony medicinal material is brittle and fibrous, so the surface of the powder is relatively smooth. The water content of the scutellaria powder sample is low and the hygroscopicity is poor, which has better stability than the other three medicinal materials. However, there are large fluctuations in the Carr index and porosity between different samples of this medicinal material, which indicates that the difference in particle size will have a huge impact on the compressibility of the sample. Therefore, when developing preparations with scutellaria as the raw medicinal material, the particle size of the medicinal powder should be strictly controlled to reduce the impact on the downstream process. Since Qingfengteng and Pueraria belong to the same rhizome medicinal materials, the physical fingerprints of the two are relatively similar. It can be seen that the fluidity of the powders of these two medicinal materials is relatively poor, but the compressibility is good. Since the water content of the powder samples of Sinomenium scutellariae, White Peony Root and Pueraria lobata is relatively high, these three medicinal materials need to be fully dried during the production process to reduce the impact of moisture.

[0048] The correlation analysis results of 11 physical parameters of 4 kinds of medicinal powders are as follows: Figure 4 As shown in the figure. For these four medicinal powders, the angle of repose is one of the most critical physical parameters, and the results show that it has a strong correlation with most physical parameters of each sample. This is consistent with the results obtained in the second chapter, that is, fluidity is a key physical property of micro-pellet cores, Chinese medicine powders and even more powder samples, which can reflect other physical properties of the samples to a certain extent.

[0049] Example 3 Model establishment and evaluation

[0050] 1. Spectral preprocessing

[0051] For each medicinal material, 30 powder samples from 10 batches of medicinal materials were randomly selected, and the spectra of these 30 samples were divided into 21 training set spectra and 9 validation set spectra using the spectral-physical and chemical value co-occurrence distance (Sample set partitioning xy distance, SPXY) algorithm. The spectra of the remaining 9 samples from the remaining 3 batches of medicinal materials were used as independent test set spectra. The SPXY algorithm ensures that the samples have the same weight in the spectral and measured variable space, which helps to build a more representative training set. The cosine similarity value of the angle between the physical fingerprint prediction spectrum and the actual spectrum of the validation set sample was used as the evaluation index, and the best method was selected from the four preprocessing methods of SNV, SG smoothing, 1st Der-SG smoothing and SNV-1st Der-SG smoothing.

[0052] The prediction results of the physical fingerprints of the validation set samples based on different preprocessing methods are shown in Table 3. According to the results, the best spectral preprocessing methods for white peony root, scutellaria baicalensis, sinomenium sinensis and pueraria root samples were finally determined to be SNV, SNV-1stDer-SG smoothing, SNV-1st Der-SG smoothing, and 1st Der-SG smoothing, respectively. Figure 3 shown.

[0053] Table 3 Cosine similarity values ​​of angles of physical fingerprint spectra of medicinal material validation set samples after different pretreatment methods

[0054]

[0055] 2. Variable screening and model building

[0056] For each medicinal material, Spearman correlation analysis was used to screen the spectral variables that were highly correlated with each physical parameter, and then integrated. Before and after variable screening, the DS, PLSR and GRNN algorithms were used to build models, and the model performance was compared. Then, peak attribution analysis was performed.

[0057] In order to reduce the complexity of the model, the DS algorithm was first used to establish a model based on the original spectral variables. The model prediction results are shown in Table 4. The prediction accuracy of the model for each physical parameter is Figure 6-Figure 9It is shown in the figure, and the dotted line in the figure is the connecting line of the average absolute error. It should be noted that when measuring the physical parameters of the validation set samples of Qingfeng Caulis and Pueraria lobata, due to the large sample consumption, there is one sample in both medicinal materials that cannot meet the measurement requirements of all physical parameters. Therefore, the number of validation set samples of Qingfeng Caulis and Pueraria lobata is 8. It can be seen from Table 4 that when using the DS model, there are samples with angle cosine similarity values ​​lower than 0.99 and Mahalanobis distance greater than 1.5 in the validation set samples of the four medicinal materials. To further optimize the model, this study used PLSR and GRNN algorithms to establish a model based on the original spectral variables. The optimization results of the key parameters of the two models are shown in Tables 5-6. Compared with the DS model, when using the PLSR and GRNN models, the angle cosine similarity values ​​of the validation set samples of the four medicinal materials are higher and the Mahalanobis distance is smaller. In addition, it can be observed from the figure that the two models have a higher prediction accuracy for the physical parameters of the samples. After comparing the PLSR model and GRNN model of the four medicinal materials, it was found that compared with the GRNN algorithm, the PLSR algorithm was more suitable for establishing the physical fingerprint conversion model of Sinomenium scabra and white peony powder, while the GRNN algorithm was more suitable for establishing the relevant model of Scutellaria baicalensis and Pueraria lobata powder.

[0058] Table 4 Model prediction results based on original spectral variables

[0059]

[0060]

[0061] Table 5 Optimization results of PLSR model parameters based on original spectral variables

[0062]

[0063]

[0064] Note: α: angle of repose; c: angle of collapse; Da: bulk density; Dc: tap density; IC: Carr index; IH: Hausner ratio; Ie: porosity between powders; %HR: loss on drying; %H: hygroscopicity; %pf: percentage of particle size <50μm; Iθ: particle size uniformity

[0065] Table 6 GRNN model parameter optimization results based on original spectral variables

[0066]

[0067] Note: α: angle of repose; c: angle of collapse; Da: bulk density; Dc: tap density; IC: Carr index; IH: Hausner ratio; Ie: porosity between powders; %HR: loss on drying; %H: hygroscopicity; %pf: percentage of particle size <50μm; Iθ: particle size uniformity

[0068] 3. Model performance evaluation indicators

[0069] This study uses two methods to evaluate the performance of the model. First, the absolute error is used to evaluate the model's prediction ability for each physical parameter. In addition, the angle cosine similarity and Mahalanobis distance are used to evaluate the similarity between the predicted spectrum of the sample physical fingerprint and the actual spectrum.

[0070] In order to investigate whether VS can improve the performance of the model, the key variables screened by Spearman correlation analysis were used to establish the model. The optimization results of the key parameters of the PLSR model and the GRNN model are shown in Tables 7-8. The prediction results of the established model are shown in Table 9. After variable screening, the performance of the PLSR model was improved to a certain extent. Although the number of samples with angle cosine similarity values ​​less than 0.99 was still 2, the number of validation set samples with Mahalanobis distance greater than 1.0 was reduced from 7 to 4. VS significantly improved the performance of the GRNN model. The number of validation set samples with angle cosine similarity values ​​less than 0.99 was reduced from 3 to 1, and the number of validation set samples with Mahalanobis distance greater than 1.0 was reduced from 7 to 3. The mean absolute error of the physical parameter prediction results showed a downward trend, indicating that the prediction accuracy of the physical parameters of the samples by the two models was improved. On the contrary, the performance of the DS model decreased significantly after VS. The reason may be that the reduction of variables led to a reduction in the information contained in the pseudo-inverse matrix, thus affecting the performance of the model. The above results show that the prediction performance of the GRNN model is the best. Fig.10 The figure shows the correlation diagram between the predicted value based on the GRNN model and the actual measured value. It can be seen from the figure that the prediction accuracy of the model is high. Fig.11 The physical fingerprint prediction map and actual map of the validation set samples based on the model are shown. The figure clearly shows the high similarity between the two, which further proves the excellent prediction ability of the model. After the development of the model was completed, the GRNN model was applied to the prediction of the physical fingerprint maps of the independent test set samples of 4 medicinal materials. The results showed that the predicted maps of the samples were highly similar to the actual maps, and the angle cosine similarity values ​​were all greater than 0.99, which shows that the model has strong generalization ability.

[0071] Table 7 Optimization results of PLSR model parameters based on key spectral variables

[0072]

[0073]

[0074] Table 8 Optimization results of GRNN model parameters based on key spectral variables

[0075]

[0076] Table 9 Model prediction results based on key spectral variables

[0077]

[0078]

[0079] Sugar and cellulose are the components with higher content in Radix Paeoniae Alba and Radix Scutellariae. Their characteristic absorption in NIR spectrum is mainly located at 4000cm -1 Nearby. From Fig.12 It can be observed in A and 12B that some key spectral variables of white peony root and scutellaria root samples are located at 4000 cm -1 This indicates that there is a correlation between the selected variables and the chemical composition. The first combined frequency region of the stretching vibration and bending vibration of the CH group is mainly located at 4500 cm -1 -4545cm -1 The absorption band between the two bands is located at 7100 cm -1 、7263cm -1 and 7355cm -1 The second harmonic frequency region of CH group stretching vibration is located at 8264cm -1 -8696cm -1 For the OH group, the second harmonic frequency region of its stretching vibration appears at 9550cm -1 In addition, the first harmonic frequency region of the stretching vibration of the NH group is at 6600cm -1 In this study, the key spectral variables screened all appear in the above characteristic absorption regions ( Fig.12 ), indicating that the constructed model has good interpretability.

[0080] 4. Simplification of the model

[0081] After screening out the optimal model, a simplified model was established using spectral variables that were highly correlated with the key physical parameters of each medicinal material powder, and the performance of the model was evaluated.

[0082] According to the results of the correlation analysis of physical parameters, 40 spectral variables that were highly correlated with the angle of repose and hygroscopicity were used to construct the GRNN simplified model. As can be seen from Table 10, when the simplified model was used, the angle cosine similarity values ​​of all samples were greater than 0.99, which indicated that variable simplification did not affect the performance of the GRNN model.

[0083] Table 10 Prediction results of validation set samples of GRNN simplified model

[0084]

[0085]

[0086] In summary, the GRNN model established by the present invention can accurately predict the physical parameters of four kinds of medicinal powders, including the Carr index and the porosity between powders, without using ensemble learning. The similarity between the predicted physical fingerprint spectrum and the actual spectrum of the sample is high. The reason for this phenomenon is that there are large differences in the physical parameters of medicinal powders in different particle size ranges, so the difference between the physical fingerprint spectra of the modeled samples is relatively large. This reflects to a certain extent that the data set has good representativeness, and the model established using representative data often has good prediction performance and generalization ability. Based on the above results, it can be found that when establishing a physical fingerprint spectrum conversion model for samples with small differences in physical parameters, it is necessary to conduct more in-depth exploration in terms of data set expansion and modeling strategy optimization. For samples with large differences in physical parameters, the modeling difficulty may be smaller, and more in-depth model simplification research can be carried out.

[0087] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for rapidly characterizing the physical parameters of traditional Chinese medicine powder, characterized in that: The steps include: (1) Determine the near infrared spectrum of traditional Chinese medicine powder samples; (2) determining the physical parameters of the Chinese medicine powder sample and constructing a physical fingerprint; (3) Preprocessing the near-infrared spectrum, screening spectral variables related to the physical parameters, establishing a spectral variable model, and characterizing the physical parameters.

2. The method for rapidly characterizing the physical properties of Chinese medicine powder according to claim 1, characterized in that: The Chinese medicine powder sample in step (1) is at least one of Caulis Sinomenii, Radix Scutellariae, Radix Puerariae or Radix Paeoniae Alba.

3. The method for rapidly characterizing the physical properties of Chinese medicine powder according to claim 1, characterized in that: The particle size of the traditional Chinese medicine powder sample is 0.1-355 μm.

4. The method for rapidly characterizing the physical properties of Chinese medicine powder according to claim 1, characterized in that: The near infrared spectrum acquisition condition in step (1) is Empty 1×-16cm -1 Resolution - 32 scans, Empty 2× - 8cm -1 Resolution - 64 scans, Empty 2× - 16cm -1 Resolution - 64 scans or Empty 2× - 8cm -1 Resolution - At least one of the 32 scans.

5. The method for rapidly characterizing the physical properties of Chinese medicine powder according to claim 4, characterized in that: The near infrared spectrum acquisition condition in step (1) is Empty 2×-8cm -1 Resolution - 64 scans.

6. The method for rapidly characterizing the physical properties of Chinese medicine powder according to claim 1, characterized in that: The physical parameter in step (2) is at least one of bulk density, tap density, angle of repose, angle of collapse, loss on drying, hygroscopicity, porosity between powders, Carr index, Hausner ratio, percentage of particles less than 50 μm, or particle size uniformity.

7. The method for rapidly characterizing the physical properties of Chinese medicine powder according to claim 1, characterized in that: The method of the preprocessing step in step (3) is at least one of SNV, SG smoothing, 1st Der-SG smoothing or SNV-1st Der-SG smoothing.

8. The method for rapidly characterizing the physical properties of Chinese medicine powder according to claim 1, characterized in that: The method for establishing the spectral variable model in step (3) is at least one of DS, PLSR or GRNN.

9. The method for rapidly characterizing the physical properties of Chinese medicine powder according to claim 8, characterized in that: The method for establishing the spectral variable model in step (3) is PLSR or GRNN.

Images