Application of intelligent manufacturing mass transfer model of traditional Chinese medicine tablets in quality control of ginkgo leaf preparations

Abstract

The present invention provides an application of a quality transfer model for intelligent manufacturing of traditional Chinese medicine tablets in the quality control of ginkgo preparations, which is characterized in that, based on physical quality attribute index data and chemical quality attribute data, a chemometric method is used to analyze the process mass transfer of traditional Chinese medicine tablet raw material preparation, granulation, and tablet pressing units, determine the feasible domain of the quality of intelligent manufacturing of traditional Chinese medicine tablets, and provide a basis for quality evaluation of traditional Chinese medicine tablets.

Description

Technical Field

[0001] The invention belongs to the field of traditional Chinese medicine, and particularly relates to the application of a mass transfer model for intelligent manufacturing of traditional Chinese medicine tablets in the quality control of ginkgo leaf preparations. Technical Background

[0002] The production process of Chinese medicine preparations is a specific stage for the formation of the inherent quality of Chinese medicine manufacturing products. Each unit link in the Chinese medicine manufacturing process will have an impact on the quality of Chinese medicine. The production process of Chinese medicine preparations is a complex system and an organic whole. The production of various dosage forms requires a series of operation units, such as pre-treatment of medicinal materials, preparation molding, packaging and storage. Products at different stages, through one or more physical, chemical and biological information transmission or integration, constitute the comprehensive quality of the product and ultimately affect the efficacy of Chinese medicine products in the human body.

[0003] Due to the particularity of Chinese medicine preparations, the quality control of Chinese medicine manufacturing products has long tended to be based on the traditional post-control mode, resulting in unclear correlation between the control level of each operating unit in the pharmaceutical process and the quality standards of intermediates / finished products, lack of effective quality standard constraints, and thus leading to defects such as large batch differences, lack of traceability, and difficulty in control. In the manufacturing process of Chinese medicine tablets, the prescription raw materials are the source of process input, and their quality variations will be transmitted through each link of the molding process, resulting in quality differences in the final product.

[0004] Ginkgo leaf preparations are important clinical cardiovascular drugs. They are made from ginkgo leaf extracts and appropriate excipients, including ginkgo leaf tablets (including dispersible tablets) and ginkgo leaf capsules, ginkgo leaf granules, ginkgo leaf drops, ginkgo leaf pills, ginkgo leaf oral liquid, ginkgo dew, ginkgo leaf soft capsules, etc. At present, ginkgo leaf drugs on the market are mainly ginkgo leaf tablets. However, ginkgo leaf extracts produced by different manufacturers have obvious differences. Based on the establishment of intelligent manufacturing mass transfer model, the mass transfer change law of material flow is studied, which is helpful to realize the digital pharmaceutical production and process quality control of ginkgo leaf tablets. Summary of the invention

[0005] In order to solve the existing quality difference problem of traditional Chinese medicine preparations mentioned above, the purpose of the present invention is to provide a mass transfer model for intelligent manufacturing of traditional Chinese medicine tablets.

[0006] Another object of the present invention is to provide an application of a mass transfer model for intelligent manufacturing of traditional Chinese medicine tablets in the quality control of ginkgo leaf preparations.

[0007] The present invention first provides a quality transfer model for intelligent manufacturing of traditional Chinese medicine tablets, including principal component analysis, multivariate statistical process control, hierarchical cluster analysis, partial least squares, orthogonal partial least squares-discrimination. On this basis, the Monte Carlo sampling method is used to repeatedly sample and model, and the feasible threshold of the quality of intelligent manufacturing of traditional Chinese medicine tablets is determined. The specific steps are as follows:

[0008] (1) Collection of physical and chemical quality attribute data of raw materials, granules, and tablets. Physical quality attribute data include 5 to 35 index data detected and calculated by instruments, and chemical quality attribute data include near-infrared spectral data;

[0009] (2) Preprocessing of physical and chemical quality attribute data of raw materials, granules, and tablets. Data preprocessing methods include mean centroid transformation, standardization, moving mean smoothing, Savitzky-Golay convolution smoothing, derivative method, standard normal variate transformation, and multivariate scatter correction;

[0010] (3) Construction of quality transfer model for intelligent manufacturing of traditional Chinese medicine tablets, including principal component analysis, multivariate statistical process control, hierarchical cluster analysis, partial least squares, and orthogonal partial least squares-discriminant;

[0011] (4) Construction of the quality transfer model for intelligent manufacturing of Chinese medicine tablets. The Monte Carlo sampling method is used to repeatedly perform sampling modeling to determine the feasible threshold of the quality of intelligent manufacturing of Chinese medicine tablets.

[0012] The present invention also provides an application of the intelligent manufacturing mass transfer model of traditional Chinese medicine tablets in the quality control of ginkgo leaf preparations, including ginkgo leaf extracts, ginkgo leaf granules, and ginkgo leaf tablets. The specific steps are as follows:

[0013] (1) The physical quality attributes of Ginkgo biloba extract, Ginkgo biloba granules, and Ginkgo biloba leaves were characterized by 5 to 35 indicators; the chemical quality attributes of Ginkgo biloba extract, Ginkgo biloba granules, and Ginkgo biloba leaves were characterized by near-infrared spectrometer, with the air inside the instrument as the background, and the spectrum was collected in diffuse reflectance mode, with the collection wavelength point of 0.5 nm and the spectral scanning range of 1100 to 2300 nm;

[0014] (2) Preprocessing of the physical and chemical quality attribute data of Ginkgo biloba extract, Ginkgo biloba granules, and Ginkgo biloba leaf tablets. The data preprocessing methods include mean centroid transformation, standardization, moving mean smoothing, Savitzky-Golay convolution smoothing, derivative method, standard normal variable transformation, and multivariate scatter correction;

[0015] (3) Construction of the quality transfer model for intelligent manufacturing of ginkgo leaves, which includes principal component analysis, multivariate statistical process control, hierarchical cluster analysis, partial least squares, and orthogonal partial least squares-discriminant;

[0016] (4) The construction of the quality transfer model for intelligent manufacturing of ginkgo leaf tablets was carried out. The distances between samples were defined as Euclidean distance, standard Euclidean distance, and Mahalanobis distance. The distances between classes were defined as the shortest distance method, the longest distance method, the middle distance method, the centroid method, the class average method, and the sum of squared deviations method. The Monte Carlo sampling method was used to randomly select 60% to 80% of the samples from the data set for hierarchical clustering analysis and modeling. The sampling modeling was repeated in this way. The feasible threshold of the quality of ginkgo leaf extract, ginkgo leaf granules, and ginkgo leaf tablets was determined with a frequency distribution of 90% to 95%.

[0017] Further preferably, the step (2) is: preprocessing the physical quality attribute data by using the mean normalization method, preprocessing the near infrared spectrum data by using the moving mean smoothing method, and the window width value is 5;

[0018] Further preferably, the step (3) is: using a method with a fitting variance greater than 95% to determine the number of principal components, establishing a principal component analysis model, determining the coupling effect between variables and the characteristic distribution of samples, and extracting characteristic components; using the Hotelling T 2 Graph and SPE graph, set 95% confidence limit to identify abnormal samples in the data;

[0019] Further preferably, the step (4) is as follows: the distance between samples is defined as the Euclidean distance, the distance between classes is defined as the sum of squared deviations, 70% of the samples are randomly selected from the data set using the Monte Carlo sampling method for hierarchical clustering analysis, and this sampling is repeated 10,000 times for modeling, and the feasible threshold of the quality of the ginkgo leaf extract, granules, and tablets is determined with a frequency distribution of 95%. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 ; Hotelling T2 and SPE plots of physical properties of Ginkgo biloba extract, granules and tablets

[0021] Figure 2 ; HCA frequency distribution and dendrogram of physical properties of Ginkgo biloba extract, granules and tablets Figure 3 ; NIR raw and pre-processed spectra of Ginkgo biloba extract, granules and tablets

[0022] Figure 4 ; Hotelling T2 and SPE plots of NIR spectra of Ginkgo biloba extract, granules and tablets

[0023] Figure 5 ; HCA frequency distribution and dendrogram of NIR spectra of Ginkgo biloba extract, granules and tablets

[0024] Figure 6; HCA frequency distribution diagram and tree diagram of information fusion of Ginkgo biloba extract, granules and tablets DETAILED DESCRIPTION

[0025] The present invention will be further described below in conjunction with the accompanying drawings and specific embodiments.

[0026] This example uses the physical quality attribute data and near-infrared spectral data of ginkgo leaf extract, intermediate particles and finished tablets in ginkgo leaf tablet preparations as research objects, and specifically explains in detail the establishment of a mass transfer model for the production process of traditional Chinese medicine tablets. These three examples do not constitute a limitation on the scope of the method of the present invention.

[0027] Example 1: Mass transfer of ginkgo leaf process based on material flow and physical mass attributes

[0028] 1) 30 batches of ginkgo biloba extracts from different sources and batches, as well as the corresponding intermediate granules and finished tablets. The collection of physical quality attribute data included but was not limited to 34 indicators of bulking, compressibility, flowability, uniformity and stability. Among them, powder was characterized by 29 indicators, granules were characterized by 28 indicators, and tablets were characterized by 5 indicators, for a total of 90 samples.

[0029] 2) The mean standardization method was used to preprocess the data to eliminate the dimensional differences between different variables. The number of principal components was determined by the method of fitting variance greater than 95%, and PCA was performed on the physical quality attribute data. The PCA results of the physical properties of the three unit links are shown in Table 1. 8, 9, and 4 principal components were selected to establish the PCA model, and the cumulative variance contribution rates were 96.6%, 96.4%, and 99.9%, respectively.

[0030] Table 1 PCA results of raw materials, granules and tablets

[0031]

[0032] 3) Hotelling T of Ginkgo biloba extract, granules and tablets 2 See the picture Figure 1 (a, b, c), SPE diagram see Figure 1 (d, e, f). The results showed that the abnormal batches of Ginkgo biloba extracts were B3 and B9; there were no abnormal batches of Ginkgo biloba granules; and the abnormal batch of Ginkgo biloba tablets was B15.

[0033] 4) The MCS technique was used to randomly sample 10,000 times, and the distance between samples was defined as the Euclidean distance, and the distance between classes was defined as the ward. Each time, 70% of the samples were randomly selected from the data set for HCA modeling. The HCA frequency distribution histogram and dendrogram of Ginkgo biloba extract, granules and tablets are shown in Figure 2(a, b, c). The feasible thresholds of the quality of the three unit products of raw materials, granules and tablets were calculated with a frequency distribution of 95%. The results showed that the physical quality attribute co-class distance range of Ginkgo biloba extract was 8.7449-10.3706, the physical quality attribute co-class distance range of Ginkgo granules was 5.9351-7.8567, and the physical quality attribute co-class distance range of Ginkgo biloba tablets was 1.9875-2.7647. This shows that some quality differences of raw materials can be gradually eliminated in the granulation and tableting unit processes, and within the above threshold range, the physical quality attributes of Ginkgo biloba extract, granules and tablets are good.

[0034] Example 2: Mass transfer of ginkgo leaf process based on material flow and NIR spectroscopy

[0035] 1) 30 batches of ginkgo leaf extracts from different sources and batches, as well as the corresponding intermediate granules and finished tablets, were collected using a near-infrared spectrometer with the air inside the instrument as the background. The spectra were collected in diffuse reflectance mode with a wavelength of 0.5 nm, a spectral scanning range of 1100-2300 nm, and 32 scans. Six samples were collected from each batch, each sample was scanned 3 times, and the average spectrum was taken. A total of 90 samples were collected.

[0036] 2) The NIR spectral data were preprocessed using the moving mean smoothing method with a window width of 5. The original NIR spectra of Ginkgo biloba extract, granules and tablets are shown in Figure 3 (a, b, c), NIR spectra after moving mean smoothing preprocessing are shown in Figure 3 (d, e, f).

[0037] 3) The number of principal components was determined by using a method with a fitting variance greater than 95%, and PCA was performed on the NIR spectral data. The PCA results of the NIR spectral information of the products of the three unit links are shown in Table 2. The first two principal components were selected to establish the PCA model, and the cumulative variance contribution rates were 98.5%, 97.7%, and 98.7%, respectively.

[0038] Table 2 PCA results of NIR spectral data of raw materials, granules and tablets

[0039]

[0040] 4) Hotelling T of Ginkgo biloba extract, granules and tablets 2 See the picture Figure 4 (a, b, c), SPE diagram see Figure 4 (d, e, f). The results showed that the abnormal batches of Ginkgo biloba extracts were A2 and A7; the abnormal batches of Ginkgo biloba granules were A4 and B18; and the abnormal batches of Ginkgo biloba tablets were B3 and B6.

[0041] 5) The MCS technique was used to randomly sample 10,000 times, and the distance between samples was defined as the Euclidean distance, and the distance between classes was defined as the ward. Each time, 70% of the samples were randomly selected from the data set for HCA modeling. The HCA frequency distribution histogram and dendrogram of Ginkgo biloba extract, granules and tablets are shown in Figure 5 (a, b, c). The feasible thresholds of the quality of the three unit products of raw materials, granules and tablets were calculated with a frequency distribution of 95%. The results showed that the physical quality attribute co-class distance range of Ginkgo biloba extract was 80.4407-102.3555, the physical quality attribute co-class distance range of Ginkgo granules was 56.0650-76.0162, and the physical quality attribute co-class distance range of Ginkgo biloba tablets was 38.6988-63.2815. It shows that some quality differences of raw materials can be gradually eliminated in the granulation and tableting unit processes, and within the above threshold range, the NIR spectral attributes of Ginkgo biloba extract, granules and tablets are good.

[0042] Example 3: Mass transfer of ginkgo leaf process based on material flow and information fusion

[0043] 1) 30 batches of ginkgo leaf extracts from different sources and batches, as well as the corresponding intermediate granules and finished tablets. The collection of physical quality attribute data included but was not limited to 34 indicators of bulking, compressibility, flowability, uniformity and stability, for a total of 90 samples.

[0044] 2) 30 batches of ginkgo leaf extracts and corresponding intermediate granules and finished tablets from different sources and batches were collected using a near-infrared spectrometer with the air inside the instrument as the background. The spectra were collected in diffuse reflectance mode with a wavelength of 0.5 nm, a spectral scanning range of 1100-2300 nm, 32 scans, 6 samples were collected from each batch, each sample was scanned 3 times, and the average spectrum was taken. A total of 90 samples.

[0045] 3) The NIR spectral data were preprocessed using the moving mean smoothing method with a window width of 5.

[0046] 4) The number of principal components was determined by using a method with a fitting variance greater than 95%, and PCA was performed on the physical quality attribute data and NIR spectrum data respectively to extract characteristic components for data fusion. The PCA results of the physical quality attributes and NIR spectrum attributes of the products of the three unit links are shown in Table 3, and the extracted characteristic component information was fused.

[0047] Table 2 Physical properties of raw materials, granules and tablets and NIR spectra PCA results

[0048]

[0049] 5) For the feature layer fusion data, the MCS technology was used to randomly sample 10,000 times, and the distance between samples was defined as the Euclidean distance, and the distance between classes was defined as ward. Each time, 70% of the samples were randomly selected from the data set for HCA modeling. The HCA frequency distribution histogram and dendrogram of Ginkgo biloba extract, granules and tablets are shown in Figure 6 (a, b, c). The feasible thresholds of the quality of the three unit products of raw materials, granules and tablets were calculated with a frequency distribution of 95%. The results showed that the physical quality attribute co-class distance range of Ginkgo biloba extract was 2.9287-3.5038, the physical quality attribute co-class distance range of Ginkgo granules was 2.4676-2.9403, and the physical quality attribute co-class distance range of Ginkgo biloba tablets was 1.4803-1.8762. This shows that some quality differences of raw materials can be gradually eliminated in the granulation and tableting unit processes, and within the above threshold range, the overall quality of Ginkgo biloba extract, granules and tablets is good.

Claims

1. Application of intelligent manufacturing quality transfer model of traditional Chinese medicine tablets in quality control of ginkgo leaf preparations, characterized in that: The specific steps are as follows: (1) The physical quality attributes of Ginkgo biloba extract, Ginkgo biloba granules, and Ginkgo biloba leaves were characterized by 5 to 35 indicators; the chemical quality attributes of Ginkgo biloba extract, Ginkgo biloba granules, and Ginkgo biloba leaves were characterized by near-infrared spectrometer, with the air inside the instrument as the background, and the spectrum was collected in diffuse reflectance mode, with the collection wavelength point of 0.5 nm and the spectral scanning range of 1100 to 2300 nm; (2) Preprocessing of the physical and chemical quality attribute data of Ginkgo biloba extract, Ginkgo biloba granules, and Ginkgo biloba leaf tablets. The data preprocessing methods include mean centroid transformation, standardization, moving mean smoothing, Savitzky-Golay convolution smoothing, derivative method, standard normal variable transformation, and multivariate scatter correction; (3) Construction of the quality transfer model for intelligent manufacturing of ginkgo leaves. The model includes principal component analysis, multivariate statistical process control, hierarchical cluster analysis, partial least squares, and orthogonal partial least squares-discriminant. In step (3), the number of principal components is determined by using a method with a fitting variance greater than 95%, a principal component analysis model is established, and characteristic components are extracted; the HotellingT in the multivariate statistical process control method is used 2 Graph and SPE graph, set 95% confidence limit to identify abnormal samples in the data; (4) The construction of the quality transfer model for intelligent manufacturing of ginkgo leaf tablets was carried out. The distances between samples were defined as Euclidean distance, standard Euclidean distance, and Mahalanobis distance. The distances between classes were defined as the shortest distance method, the longest distance method, the middle distance method, the centroid method, the class average method, and the sum of squared deviations method. The Monte Carlo sampling method was used to randomly select 60% to 80% of the samples from the data set for hierarchical clustering analysis and modeling. The sampling modeling was repeated in this way. The feasible threshold of the quality of ginkgo leaf extract, ginkgo leaf granules, and ginkgo leaf tablets was determined with a frequency distribution of 90% to 95%.

2. The application of the intelligent manufacturing mass transfer model of traditional Chinese medicine tablets according to claim 1 in the quality control of ginkgo leaf preparations is characterized in that: In step (2), the physical quality attribute data are preprocessed by the mean normalization method; the near-infrared spectral data are preprocessed by the moving mean smoothing method, and the window width value is 5.

3. The application of the intelligent manufacturing mass transfer model of traditional Chinese medicine tablets according to claim 1 in the quality control of ginkgo leaf preparations is characterized in that: In step (4), the distance between samples is defined as the Euclidean distance, and the distance between classes is defined as the sum of squared deviations. The Monte Carlo sampling method is used to randomly select 70% of the samples from the data set for hierarchical clustering analysis. This sampling modeling is repeated 10,000 times, and the feasible threshold of the quality of Ginkgo biloba extract, granules, and tablets is determined with a frequency distribution of 95%.

Images