A whole-process quality detection method of Shuangshitonglin capsules

Abstract

The application discloses a kind of whole-process quality detection methods of Shuangshi Tonglin Capsules.The method comprises the following steps: (1) establishing near-infrared spectroscopy quantitative calibration model of raw material medicine, intermediate and finished product of Shuangshi Tonglin Capsules respectively; (2) collecting near-infrared spectrum of the sample to be measured; (3) inputting the spectrum into the corresponding calibration model, and calculating the quality parameters of the sample to be measured.The raw material medicine model is used to quickly detect the content of index components, moisture and color in Phellodendri Chinensis Cortex and Salvia Miltiorrhiza; the intermediate model is used to monitor the content of active ingredients in the key process steps such as extraction and concentration online; the finished product model is used to simultaneously determine the content of verbascoside, salvianolic acid B, berberine hydrochloride, tanshinone IIA, and moisture and color in the capsule content.The application constructs a whole-process quality control system covering "raw material-process-finished product", has the advantages of rapidity, non-destructiveness, greenness and online monitoring, and significantly improves the quality control level and production efficiency of Shuangshi Tonglin Capsules.

Description

Technical Field

[0001] This invention belongs to the field of traditional Chinese medicine quality control technology, specifically involving a whole-process quality detection method for Shuangshitonglin capsules based on near-infrared spectroscopy technology, and particularly involving a whole-process near-infrared spectroscopy quality detection system covering medicinal raw materials, production intermediates to finished products. Background Technology

[0002] Shuangshi Tonglin Capsules are the core product of Shaanxi Momeide Qi Xuehe Pharmaceutical Co., Ltd. This traditional Chinese medicine compound preparation possesses the effects of clearing heat and dampness, resolving turbidity and promoting urination. Clinically, it is mainly used for chronic prostatitis belonging to the category of damp-heat obstruction syndrome. Its clinical efficacy is closely related to the quality of the preparation. Numerous clinical and pharmacological studies have shown that this preparation not only significantly improves clinical symptoms such as urinary frequency, urgency, and difficulty in urination, but also has anti-inflammatory, analgesic, and diuretic effects. Furthermore, it improves prostate tissue fibrosis and hyperplasia by inhibiting the TGF-β1 / Smad pathway, and exhibits synergistic effects when used in combination with antibiotics. This preparation is composed of ten traditional Chinese medicines, including Phellodendron bark, Salvia miltiorrhiza, Dioscorea hypoglauca, Patrinia scabiosaefolia, Indigo naturalis, Talcum, Plantago asiatica, Acorus tatarinowii, Poria cocos, and Atractylodes lancea. The production process involves extraction, concentration, and drying.

[0003] At present, the quality control of Shuangshitonglin capsules mainly has the following problems: (1) Low efficiency of raw material inspection: The current quality control system for Chinese herbal medicine slices mainly relies on traditional analytical methods such as high performance liquid chromatography (HPLC) and gas chromatography-mass spectrometry (GC-MS). Although the accuracy is good, there are limitations such as long detection cycle, complicated operation, high cost and destructive detection, which are difficult to meet the real-time requirements of modern large-scale production; (2) Lack of online monitoring of production process: The existing production quality control adopts the terminal detection mode, which can only conduct post-process inspection of each process link, and cannot realize online real-time monitoring and feedback adjustment of key processes such as extraction and concentration; (3) Time-consuming quality inspection of finished product: The determination of the content of multiple index components of finished capsules requires complicated sample pretreatment and chromatographic analysis. The detection of a single batch takes a long time, which affects the production efficiency.

[0004] Near-infrared spectroscopy (NIRS) is a rapid, non-destructive, and environmentally friendly analytical technique with significant advantages such as fast analysis speed, no sample pretreatment required, and online monitoring capabilities, demonstrating great application potential in the field of traditional Chinese medicine. However, there are currently no reports on NIRS quality control studies for Shuangshitonglin capsules, and a systematic quality control method covering the entire "raw material-process-finished product" chain is particularly lacking.

[0005] Therefore, developing a whole-process quality detection method for Shuangshitonglin capsules based on near-infrared spectroscopy is of great significance for improving the quality control level of this product and realizing the modernization of traditional Chinese medicine compound preparation production. Summary of the Invention

[0006] The purpose of this invention is to overcome the shortcomings of the prior art and provide a whole-process quality testing method and system for Shuangshitonglin capsules. Based on near-infrared spectroscopy technology, a complete quality testing system covering three levels—raw materials, production intermediates, and finished products—is constructed to achieve full-process quality control of Shuangshitonglin capsules from incoming inspection and process monitoring to finished product release.

[0007] To achieve the above-mentioned objectives, the present invention adopts the following technical solution: On the one hand, the present invention provides a whole-process quality testing method for Shuangshitonglin capsules, including the following steps: (1) Establish a quantitative calibration model for near-infrared spectroscopy of Shuangshitonglin capsules throughout the entire process. The model includes three main categories: raw material calibration model, intermediate calibration model during production, and finished product calibration model. (2) Collect near-infrared spectra of the samples to be tested separately: Collect near-infrared spectra of the raw material sample, intermediate sample and finished capsule contents sample of Shuangshitonglin capsules to be tested separately under the same spectral acquisition conditions. (3) Model prediction of quality parameters: Input the near-infrared spectra collected in step (2) into the corresponding calibration model in step (1) and calculate the quality parameters of each sample to be tested through the model.

[0008] Furthermore, the active pharmaceutical ingredient calibration model includes two sub-models: the Phellodendron amurense calibration model and the Salvia miltiorrhiza calibration model. Guanhuangbai calibration model: Guanhuangbai medicinal material is used as the test object, and its quality parameters include four indicators: palmatine hydrochloride content, berberine hydrochloride content, moisture and color. The Salvia miltiorrhiza calibration model uses Salvia miltiorrhiza as the test object, and its quality parameters include seven indicators: rosmarinic acid content, salvianolic acid B content, cryptotanshinone content, tanshinone I content, tanshinone IIA content, moisture, and color.

[0009] Furthermore, the intermediate correction model in the production process includes six sub-models, covering the two key processes of extraction and concentration: Aqueous extract model and concentrated extract model of Phellodendron amurense: The test objects are aqueous extract and concentrated extract of Phellodendron amurense, and the quality parameters are the contents of palmatine hydrochloride and berberine hydrochloride; Tanshinone ethanol extract model and Tanshinone concentrate model: The test objects are Tanshinone ethanol extract and concentrate, and the quality parameters are the contents of rosmarinic acid, salvianolic acid B, cryptotanshinone, tanshinone I and tanshinone IIA; Model of mixed aqueous extract of six Chinese herbal medicines and model of mixed concentrated liquid of six Chinese herbal medicines: The test objects are mixed aqueous extract and concentrated liquid composed of Piper longum, Patrinia scabiosaefolia, Plantago asiatica, Talcum, Poria cocos and Salvia miltiorrhiza in the prescription ratio. The quality parameters are the contents of genipin, chlorogenic acid, verbascoside and salvianolic acid B.

[0010] Furthermore, the finished product calibration model uses the contents of Shuangshitonglin capsules as the detection object, and its quality parameters include six indicators: verbascoside content, salvianolic acid B content, berberine hydrochloride content, tanshinone IIA content, moisture, and color.

[0011] Furthermore, the uniform acquisition conditions for the near-infrared spectrum are: spectral scanning range of 10000~4000 cm⁻¹. -1 8 cm resolution -1 Each sample was scanned 2-3 times, with each scan consisting of 32 passes.

[0012] Furthermore, the method for establishing the calibration model in step (1) includes the following sub-steps: (a) Collect representative samples and determine reference values: Collect representative samples covering different origins, batches, and concentration ranges, and determine the reference values ​​of the quality parameters of each sample using a methodologically validated high-performance liquid chromatography or conventional physicochemical analysis method. (b) Acquiring near-infrared spectra: Acquire the raw near-infrared spectra of each sample according to the spectral acquisition conditions; (c) Spectral preprocessing: The acquired raw spectra are preprocessed to eliminate interference such as baseline drift, scattering effects, and instrument noise. The preprocessing method is selected from one or more combinations of Standard Normal Variable Transform (SNV), Multivariate Scattering Correction (MSC), First Derivative (1stD), Second Derivative (2stD), Norris Derivative Smoothing (ND), and Savitzky-Golay Smoothing (SG). (d) Establishing a quantitative calibration model: The spectral modeling interval was selected by analysis of variance, the number of principal factors was determined by the root mean square error of cross-validation (RMSECV) curve inflection point method, and the partial least squares (PLS) method was used to establish a quantitative calibration model between near-infrared spectra and mass parameter reference values.

[0013] Furthermore, the spectral preprocessing methods, after systematic optimization and screening, are as follows, specifically for different detection objects and indicators: For the model of berberine hydrochloride in Phellodendron amurense raw material, the Constant + 1st D + SG pretreatment method was used; for the model of palmatine hydrochloride in Phellodendron amurense raw material, the Constant + 2st D + SG pretreatment method was used; for the model of color in Phellodendron amurense raw material, the Constant + 1st D + SG pretreatment method was used; for the model of moisture in Phellodendron amurense raw material, the Constant + 2st D + ND pretreatment method was used. For the models of rosmarinic acid and salvianolic acid B in tanshinone raw material, the SNV+2stD+ND pretreatment method was used; for the model of tanshinone IIA in tanshinone raw material, the MSC+1stD+ND pretreatment method was used; for the model of cryptotanshinone in tanshinone raw material, the MSC+1stD+SG pretreatment method was used; for the model of tanshinone I in tanshinone raw material, the MSC+2stD+SG pretreatment method was used; and for the models of moisture and color in tanshinone raw material, the SNV pretreatment method was used. For the models of berberine hydrochloride and palmatine hydrochloride in the aqueous extract of Phellodendron chinense, the MSC+2stD+ND pretreatment method was used; for the models of palmatine hydrochloride and berberine hydrochloride in the concentrated extract of Phellodendron chinense, the Constant+2stD+ND pretreatment method was used. For the models of each component in the ethanol extract of tanshinone, the MSC + 2stD + ND pretreatment method was used; for the model of salvianolic acid B in the tanshinone concentrate, the MSC pretreatment method was used; for the model of tanshinone IIA in the tanshinone concentrate, the MSC + 1stD + ND pretreatment method was used; for the models of rosmarinic acid, cryptotanshinone, and tanshinone I in the tanshinone concentrate, the SNV pretreatment method was used. For the model of salvianolic acid B in the mixed aqueous extract of six Chinese herbal medicines, the MSC + 1stD + ND pretreatment method was used; for the models of genipin and verbascoside in the mixed aqueous extract of six Chinese herbal medicines, the MSC pretreatment method was used; for the model of chlorogenic acid in the mixed aqueous extract of six Chinese herbal medicines, the MSC + 2stD + ND pretreatment method was used; for the models of verbascoside and salvianolic acid B in the mixed concentrated extract of six Chinese herbal medicines, the Constant + 1stD pretreatment method was used; for the model of genipin in the mixed concentrated extract of six Chinese herbal medicines, the MSC + 1stD pretreatment method was used; for the model of chlorogenic acid in the mixed concentrated extract of six Chinese herbal medicines, the MSC + 2stD + SG pretreatment method was used. For the models of verbascoside and salvianolic acid B in the finished product of Shuangshitonglin capsules, the SNV+1stD pretreatment method was used; for the model of berberine hydrochloride in the finished product, the MSC+2stD+SG pretreatment method was used; for the model of tanshinone IIA in the finished product, the MSC+1stD pretreatment method was used; for the model of moisture in the finished product, the Constant+1stD+ND pretreatment method was used; and for the model of color in the finished product, the Constant+1stD pretreatment method was used.

[0014] On the other hand, the present invention also provides a system for implementing the above-described full-process quality inspection method, comprising: Near-infrared spectroscopy acquisition module: used to acquire near-infrared spectra of active pharmaceutical ingredient samples, intermediate samples, and finished product samples according to preset acquisition conditions; Storage module: It stores the active pharmaceutical ingredient calibration model, intermediate calibration model and finished product calibration model that have been pre-established and verified according to the method described in this invention; Calculation module: Used to input the collected spectrum of the sample to be tested into the corresponding calibration model, and to calculate the mass parameters of the sample to be tested through the model; Output module: Used to output the calculated quality parameter test results.

[0015] Compared with the prior art, the present invention has the following beneficial effects: (1) Full process coverage and strong system: For the first time, a near-infrared spectroscopy quality detection system covering the entire process of Shuangshitonglin capsules from raw materials to intermediates to finished products has been constructed, which solves the shortcomings of traditional methods that can only perform single-stage and offline detection, and realizes full-process quality traceability and control.

[0016] (2) Fast and efficient, greatly improving detection efficiency: Samples require no or only minimal pretreatment, and the analysis time for a single sample can be shortened to within 2 minutes, reducing the laboratory testing work that takes several hours to complete instantly, significantly improving production and release efficiency.

[0017] (3) Non-destructive and green, in line with sustainable development: The detection process does not consume organic solvents and does not damage the sample. The raw material can still be put into production after detection, which is both environmentally friendly and reduces costs.

[0018] (4) Simultaneous analysis of multiple indicators: A near-infrared spectral model can simultaneously predict multiple key quality parameters (such as the content of multiple active ingredients, moisture, and color), which is more in line with the characteristics of the synergistic effect of multiple components in traditional Chinese medicine compound.

[0019] (5) High prediction accuracy and good robustness: By optimizing the spectral preprocessing method and modeling parameters in a targeted manner, the established models of each link have high coefficients of determination (Rc and Rp are both greater than 0.85, and most are greater than 0.9) and low prediction errors (RMSEC and RMSEP are both small), which meet the requirements of quantitative analysis in the pharmaceutical industry. Attached Figure Description

[0020] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0021] Figure 1 shows the HPLC specificity test results of Phellodendron amurense (where: A is blank solvent; B is mixed reference standard; C is Phellodendron amurense test sample; 1 is palmatine hydrochloride, 2 is berberine hydrochloride). Figure 2 shows the original near-infrared spectra of different batches of Phellodendron amurense (where: A 230901; B 241101; C 241001; D 240501; E 240801; F 240601; G 231201; H 241101; I 241001; J 2411012; K 10 different batches). Figure 3 shows the Press diagrams of moisture content, color, and two components of Phellodendron amurense (where: A is palmatine hydrochloride; B is berberine hydrochloride; C is moisture content of Phellodendron amurense; D is color of Phellodendron amurense). Figure 4 shows the correlation between the NIRS model reference values ​​and predicted values ​​of two components, moisture, and color of Phellodendron amurense from different batches (where: A is palmatine hydrochloride; B is berberine hydrochloride; C is moisture of Phellodendron amurense; D is color of Phellodendron amurense). Figure 5 shows the results of cross-validation of two components, moisture content, and color in different batches of Phellodendron amurense (where: A is palmatine hydrochloride; B is berberine hydrochloride; C is moisture content of Phellodendron amurense; and D is color of Phellodendron amurense). Figure 6 shows the analysis results of the measured values ​​and predicted values ​​of the two components, moisture and color in the validation set (where: A is palmatine hydrochloride; B is berberine hydrochloride; C is the moisture content of Phellodendron amurense; D is the color of Phellodendron amurense). Figure 7 shows the HPLC specificity test results of Salvia miltiorrhiza (where: A is blank solvent; B is mixed reference standard; C is tanshinone alcohol extract-concentrate test sample; 1 is rosmarinic acid; 2 is salvianolic acid B; 3 is cryptotanshinone; 4 is tanshinone I; 5 is tanshinone IIA). Figure 8 shows the original near-infrared spectra of Salvia miltiorrhiza from different producing areas (where: A Anhui; B Gansu; C Hebei; D Henan; E Shandong; F Shanxi; G Shaanxi; H Sichuan; I Yunnan; J 9 producing areas). Figure 9 shows the moisture content, color, and five component correction set Press diagrams of Salvia miltiorrhiza from different origins (where: A is rosmarinic acid; B is salvianolic acid B; C is cryptotanshinone; D is tanshinone I; E is tanshinone IIA; F is moisture content of Salvia miltiorrhiza; G is color of Salvia miltiorrhiza). Figure 10 shows the correlation between the NIRS model reference values ​​and predicted values ​​of five components, moisture content, and color of Salvia miltiorrhiza from different origins (where: A is rosmarinic acid; B is salvianolic acid B; C is cryptotanshinone; D is tanshinone I; E is tanshinone IIA; F is moisture content of Salvia miltiorrhiza; G is color content of Salvia miltiorrhiza). Figure 11 shows the results of cross-validation of five components, moisture content, and color of Salvia miltiorrhiza from different origins (where: A is rosmarinic acid; B is salvianolic acid B; C is cryptotanshinone; D is tanshinone I; E is tanshinone IIA; F is moisture content of Salvia miltiorrhiza; G is color content of Salvia miltiorrhiza). Figure 12 shows the analysis results of the measured values ​​and predicted values ​​of the five components, moisture content, and color in the validation set (where: A is rosmarinic acid; B is salvianolic acid B; C is cryptotanshinone; D is tanshinone I; E is tanshinone IIA; F is tanshinone moisture content; G is tanshinone color). Figure 13 is a production process flow chart of Shuangshitonglin Capsules (where the red pentagrams are the sampling points for content monitoring of key processes in the production of Shuangshitonglin Capsules). Figure 14 shows the HPLC chromatogram of the aqueous extract-concentrate of Phellodendron amurense (where: A is blank solvent; B is mixed reference standard; C is the aqueous extract-concentrate of Phellodendron amurense test sample; 1 is palmatine hydrochloride; 2 is berberine hydrochloride). Figure 15 shows the original near-infrared spectra of the aqueous extract and concentrate of Phellodendron amurense (where: A is the aqueous extract of Phellodendron amurense; B is the concentrate of Phellodendron amurense). Figure 16 shows the correction set Press diagram of two components in the aqueous extract-concentrate of Phellodendron amurense (where: A is the aqueous extract-palmatine hydrochloride; B is the aqueous extract-berberine hydrochloride; C is the concentrate-palmatine hydrochloride; and D is the concentrate-berberine hydrochloride). Figure 17 shows the correlation between the near-infrared model reference values ​​and predicted values ​​of two components in the aqueous extract-concentrate of Phellodendron amurense (where: A is aqueous extract-berberine hydrochloride; B is aqueous extract-palmatine hydrochloride; C is concentrated-berberine hydrochloride; D is concentrated-palmatine hydrochloride). Figure 18 shows the results of cross-validation of two components in the aqueous extract-concentrate of Phellodendron amurense (where: A is berberine hydrochloride extracted; B is palmatine hydrochloride extracted; C is berberine hydrochloride concentrated; and D is palmatine hydrochloride concentrated). Figure 19 shows the analysis results of the measured values ​​and predicted values ​​of the two components in the validation set (where: A is water-extracted berberine hydrochloride; B is water-extracted palmatine hydrochloride; C is concentrated berberine hydrochloride; D is concentrated palmatine hydrochloride). Figure 20 shows the HPLC chromatogram of the tanshinone alcohol extract-concentrate (where: A is blank solvent; B is mixed reference standard; C is tanshinone alcohol extract-concentrate test sample; 1 is rosmarinic acid; 2 is salvianolic acid B; 3 is cryptotanshinone; 4 is tanshinone I; 5 is tanshinone IIA). Figure 21 shows the original near-infrared spectrum of the tanshinone alcohol extract-concentrate (where: A is the tanshinone alcohol extract; B is the tanshinone concentrate). Figure 22 shows the Press diagram of the calibration set of five components of the ethanol extract-concentrate of Tanshinone (where: A is ethanol extract-rosmarinic acid; B is ethanol extract-tanshinone B; C is ethanol extract-cryptotanshinone; D is ethanol extract-tanshinone I; E is ethanol extract-tanshinone IIA; F is concentrate-rosmarinic acid; G is concentrate-tanshinone B; H is concentrate-cryptotanshinone; I is concentrate-tanshinone I; J is concentrate-tanshinone IIA). Figure 23 shows the correlation between near-infrared model reference values ​​and predicted values ​​of five components in the ethanol extract-concentrate of Tanshinone (where: A is ethanol extract-rosmarinic acid; B is ethanol extract-tanshinone B; C is ethanol extract-cryptotanshinone; D is ethanol extract-tanshinone I; E is ethanol extract-tanshinone IIA; F is concentrate-rosmarinic acid; G is concentrate-tanshinone B; H is concentrate-cryptotanshinone; I is concentrate-tanshinone I; J is concentrate-tanshinone IIA). Figure 24 shows the results of cross-validation of five components in the ethanol extract-concentrate of Tanshinone (A: rosmarinic acid; B: salvianolic acid B; C: cryptotanshinone; D: tanshinone I; E: tanshinone IIA; F: rosmarinic acid; G: salvianolic acid B; H: cryptotanshinone; I: tanshinone I; J: tanshinone IIA). Figure 25 shows the analysis results of the measured values ​​and predicted values ​​of the five components in the validation set (where: A is rosmarinic acid extracted by alcohol; B is salvianolic acid B extracted by alcohol; C is cryptotanshinone extracted by alcohol; D is tanshinone I extracted by alcohol; E is tanshinone IIA extracted by alcohol; F is rosmarinic acid concentrated by alcohol; G is salvianolic acid B concentrated by alcohol; H is cryptotanshinone concentrated by alcohol; I is tanshinone I concentrated by alcohol; J is tanshinone IIA concentrated by alcohol). Figure 26 shows the HPLC chromatogram of the mixed water extract-concentrate (where: A is blank solvent; B is mixed reference standard; C is mixed water extract-concentrate test sample; 1 is genipinic acid; 2 is chlorogenic acid; 3 is verbascoside; 4 is salvianolic acid B). Figure 27 shows the original near-infrared spectrum of the mixed water extract-concentrate (where: A is the mixed water extract; B is the mixed concentrate). Figure 28 shows the Press diagram of the calibration set of four components in the mixed water extract-concentrate (where: A is mixed water extract-genipin; B is mixed water extract-chlorogenic acid; C is mixed water extract-verbascoside; D is mixed water extract-tanshinone B; E is mixed concentrate-genipin; F is mixed concentrate-chlorogenic acid; G is mixed concentrate-verbascoside; H is mixed concentrate-tanshinone B). Figure 29 shows the correlation between near-infrared model reference values ​​and predicted values ​​of four components in the mixed water extract-concentrate (where: A is mixed water extract-genipin; B is mixed water extract-chlorogenic acid; C is mixed water extract-verbascoside; D is mixed water extract-tanshinone B; E is mixed concentrate-genipin; F is mixed concentrate-chlorogenic acid; G is mixed concentrate-verbascoside; H is mixed concentrate-tanshinone B). Figure 30 shows the results of cross-validation of the four components of the mixed water extract-concentrate (where: A is mixed water extract-genipin; B is mixed water extract-chlorogenic acid; C is mixed water extract-verbascoside; D is mixed water extract-tanshinone B; E is mixed concentrate-genipin; F is mixed concentrate-chlorogenic acid; G is mixed concentrate-verbascoside; H is mixed concentrate-tanshinone B). Figure 31 shows the analysis results of the measured values ​​and predicted values ​​of four components in the mixed water extract-concentrate validation set (where: A is mixed water extract-genipin; B is mixed water extract-chlorogenic acid; C is mixed water extract-verbascoside; D is mixed water extract-tanshinone B; E is mixed concentrate-genipin; F is mixed concentrate-chlorogenic acid; G is mixed concentrate-verbascoside; H is mixed concentrate-tanshinone B). Figure 32 shows the HPLC chromatogram of Shuangshitonglin Capsules (where: A is the mixed reference standard; B is the Shuangshitonglin Capsule test sample; C is the blank solvent; D is the negative test sample of Danshen; E is the negative test sample of Plantago asiatica; F is the negative test sample of Phellodendron chinense; 1 is verbascoside; 2 is salvianolic acid B; 3 is berberine hydrochloride; 4 is tanshinone IIA). Figure 33 shows the original near-infrared spectrum of Shuangshitonglin capsules; Figure 34 shows the Press diagram of the calibration set of two indicators for four components of Shuangshitonglin capsules (where: A is verbascoside; B is salvianolic acid B; C is berberine hydrochloride; D is tanshinone IIA; E is the moisture content of Shuangshitonglin capsules; and F is the color value of Shuangshitonglin capsules). Figure 35 shows the correlation between the near-infrared model reference values ​​and predicted values ​​of two indicators for four components of Shuangshitonglin capsules (where: A is verbascoside; B is salvianolic acid B; C is berberine hydrochloride; D is tanshinone IIA; E is the moisture content of Shuangshitonglin capsules; and F is the color value of Shuangshitonglin capsules). Figure 36 shows the results of cross-validation of four components and two indicators of Shuangshitonglin capsules (where: A is verbascoside; B is salvianolic acid B; C is berberine hydrochloride; D is tanshinone IIA; E is the moisture content of Shuangshitonglin capsules; and F is the color value of Shuangshitonglin capsules). Figure 37 shows the analysis results of the measured values ​​and predicted values ​​of the four components, moisture, and color in the validation set. Among them: A. Verbascoside; B. Tanshinone B; C. Berberine Hydrochloride; D. Tanshinone IIA; E. Moisture content of Shuangshitonglin Capsules; F. Color value of Shuangshitonglin Capsules. Detailed Implementation

[0022] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments, but the embodiments of the present invention are not limited thereto. Other methods for preparing the compounds of the present invention, with some conventional modifications to the reaction conditions according to the present invention, are considered to be within the scope of the present invention.

[0023] Unless otherwise defined, the technical or scientific terms used in this invention shall have the ordinary meaning as understood by one of ordinary skill in the art to which this invention pertains.

[0024] On one hand, the core technical solution of this invention is: a whole-process quality testing method for Shuangshitonglin capsules, comprising the following steps: (1) Establish a quantitative calibration model for near-infrared spectroscopy of Shuangshitonglin capsules throughout the entire process. The model includes three main categories: raw material calibration model, intermediate calibration model during production, and finished product calibration model. (2) Collect near-infrared spectra of the samples to be tested separately: Collect near-infrared spectra of the raw material sample, intermediate sample and finished capsule contents sample of Shuangshitonglin capsules to be tested separately under the same spectral acquisition conditions. (3) Model prediction of quality parameters: Input the near-infrared spectra collected in step (2) into the corresponding calibration model in step (1) and calculate the quality parameters of each sample to be tested through the model.

[0025] Furthermore, the active pharmaceutical ingredient calibration model includes two sub-models: the Phellodendron amurense calibration model and the Salvia miltiorrhiza calibration model. Guanhuangbai calibration model: Guanhuangbai medicinal material is used as the test object, and its quality parameters include four indicators: palmatine hydrochloride content, berberine hydrochloride content, moisture and color. The Salvia miltiorrhiza calibration model uses Salvia miltiorrhiza as the test object, and its quality parameters include seven indicators: rosmarinic acid content, salvianolic acid B content, cryptotanshinone content, tanshinone I content, tanshinone IIA content, moisture, and color.

[0026] Furthermore, the intermediate correction model in the production process includes six sub-models, covering the two key processes of extraction and concentration: Aqueous extract model and concentrated extract model of Phellodendron amurense: The test objects are aqueous extract and concentrated extract of Phellodendron amurense, and the quality parameters are the contents of palmatine hydrochloride and berberine hydrochloride; Tanshinone ethanol extract model and Tanshinone concentrate model: The test objects are Tanshinone ethanol extract and concentrate, and the quality parameters are the contents of rosmarinic acid, salvianolic acid B, cryptotanshinone, tanshinone I and tanshinone IIA; Model of mixed aqueous extract of six Chinese herbal medicines and model of mixed concentrated liquid of six Chinese herbal medicines: The test objects are mixed aqueous extract and concentrated liquid composed of Piper longum, Patrinia scabiosaefolia, Plantago asiatica, Talcum, Poria cocos and Salvia miltiorrhiza in the prescription ratio. The quality parameters are the contents of genipin, chlorogenic acid, verbascoside and salvianolic acid B.

[0027] Furthermore, the finished product calibration model uses the contents of Shuangshitonglin capsules as the detection object, and its quality parameters include six indicators: verbascoside content, salvianolic acid B content, berberine hydrochloride content, tanshinone IIA content, moisture, and color.

[0028] Furthermore, the uniform acquisition conditions for the near-infrared spectrum are: spectral scanning range of 10000~4000 cm⁻¹. -1 8 cm resolution -1 Each sample was scanned 2-3 times, with each scan consisting of 32 passes.

[0029] Furthermore, the method for establishing the calibration model in step (1) includes the following sub-steps: (a) Collect representative samples and determine reference values: Collect representative samples covering different origins, batches, and concentration ranges, and determine the reference values ​​of the quality parameters of each sample using a methodologically validated high-performance liquid chromatography or conventional physicochemical analysis method. (b) Acquiring near-infrared spectra: Acquire the raw near-infrared spectra of each sample according to the spectral acquisition conditions; (c) Spectral preprocessing: The acquired raw spectra are preprocessed to eliminate interference such as baseline drift, scattering effects, and instrument noise. The preprocessing method is selected from one or more combinations of Standard Normal Variable Transform (SNV), Multivariate Scattering Correction (MSC), First Derivative (1stD), Second Derivative (2stD), Norris Derivative Smoothing (ND), and Savitzky-Golay Smoothing (SG). (d) Establishing a quantitative calibration model: The spectral modeling interval was selected by analysis of variance, the number of principal factors was determined by the root mean square error of cross-validation (RMSECV) curve inflection point method, and the partial least squares (PLS) method was used to establish a quantitative calibration model between near-infrared spectra and mass parameter reference values.

[0030] Furthermore, the spectral preprocessing methods, after systematic optimization and screening, are as follows, specifically for different detection objects and indicators: For the model of berberine hydrochloride in Phellodendron amurense raw material, the Constant + 1st D + SG pretreatment method was used; for the model of palmatine hydrochloride in Phellodendron amurense raw material, the Constant + 2st D + SG pretreatment method was used; for the model of color in Phellodendron amurense raw material, the Constant + 1st D + SG pretreatment method was used; for the model of moisture in Phellodendron amurense raw material, the Constant + 2st D + ND pretreatment method was used. For the models of rosmarinic acid and salvianolic acid B in tanshinone raw material, the SNV+2stD+ND pretreatment method was used; for the model of tanshinone IIA in tanshinone raw material, the MSC+1stD+ND pretreatment method was used; for the model of cryptotanshinone in tanshinone raw material, the MSC+1stD+SG pretreatment method was used; for the model of tanshinone I in tanshinone raw material, the MSC+2stD+SG pretreatment method was used; and for the models of moisture and color in tanshinone raw material, the SNV pretreatment method was used. For the models of berberine hydrochloride and palmatine hydrochloride in the aqueous extract of Phellodendron chinense, the MSC+2stD+ND pretreatment method was used; for the models of palmatine hydrochloride and berberine hydrochloride in the concentrated extract of Phellodendron chinense, the Constant+2stD+ND pretreatment method was used. For the models of each component in the ethanol extract of tanshinone, the MSC + 2stD + ND pretreatment method was used; for the model of salvianolic acid B in the tanshinone concentrate, the MSC pretreatment method was used; for the model of tanshinone IIA in the tanshinone concentrate, the MSC + 1stD + ND pretreatment method was used; for the models of rosmarinic acid, cryptotanshinone, and tanshinone I in the tanshinone concentrate, the SNV pretreatment method was used. For the model of salvianolic acid B in the mixed aqueous extract of six Chinese herbal medicines, the MSC + 1stD + ND pretreatment method was used; for the models of genipin and verbascoside in the mixed aqueous extract of six Chinese herbal medicines, the MSC pretreatment method was used; for the model of chlorogenic acid in the mixed aqueous extract of six Chinese herbal medicines, the MSC + 2stD + ND pretreatment method was used; for the models of verbascoside and salvianolic acid B in the mixed concentrated extract of six Chinese herbal medicines, the Constant + 1stD pretreatment method was used; for the model of genipin in the mixed concentrated extract of six Chinese herbal medicines, the MSC + 1stD pretreatment method was used; for the model of chlorogenic acid in the mixed concentrated extract of six Chinese herbal medicines, the MSC + 2stD + SG pretreatment method was used. For the models of verbascoside and salvianolic acid B in the finished product of Shuangshitonglin capsules, the SNV+1stD pretreatment method was used; for the model of berberine hydrochloride in the finished product, the MSC+2stD+SG pretreatment method was used; for the model of tanshinone IIA in the finished product, the MSC+1stD pretreatment method was used; for the model of moisture in the finished product, the Constant+1stD+ND pretreatment method was used; and for the model of color in the finished product, the Constant+1stD pretreatment method was used.

[0031] Furthermore, the method also includes model validation and evaluation steps: the established calibration model is comprehensively evaluated using leave-one-out cross-validation (LOO-CV) and independent validation sets. The evaluation metrics include calibration set coefficient of determination (Rc), validation set coefficient of determination (Rp), calibration set root mean square error (RMSEC), and validation set root mean square error (RMSEP). All models must satisfy that Rc and Rp are both greater than 0.85, and RMSEC and RMSEP are both less than 50% of the standard deviation of the corresponding quality parameter reference values, to ensure that the model prediction accuracy meets the quality control requirements.

[0032] Furthermore, the full-process quality inspection described in this invention covers the entire production chain of Shuangshitonglin capsules, specifically including: the incoming inspection of raw materials, the online monitoring of the extraction process, the online monitoring of the concentration process, and the outgoing inspection of finished capsules.

[0033] Secondly, the present invention also provides a system for implementing the above-described full-process quality inspection method, comprising: Near-infrared spectroscopy acquisition module: used to acquire near-infrared spectra of active pharmaceutical ingredient samples, intermediate samples, and finished product samples according to preset acquisition conditions; Storage module: It stores the active pharmaceutical ingredient calibration model, intermediate calibration model and finished product calibration model that have been pre-established and verified according to the method described in this invention; Calculation module: Used to input the collected spectrum of the sample to be tested into the corresponding calibration model, and to calculate the mass parameters of the sample to be tested through the model; Output module: Used to output the calculated quality parameter test results.

[0034] The present invention will be further illustrated below by way of examples. Unless otherwise specified, the materials in the examples are prepared according to existing methods or purchased directly from the market.

[0035] Instruments and Materials Table 1 Experimental Apparatus The prepared slices of Phellodendron bark and Salvia miltiorrhiza were identified by the Department of Pharmacy of Students at Shaanxi University of Traditional Chinese Medicine as being from the Rutaceae family, specifically Phellodendron bark (…). PhellodendronamurenseThe dried bark of Rupr and the Lamiaceae plant Salvia miltiorrhiza (Rupr) Salvia miltiorrhiza The dried roots and rhizomes of Bge.

[0036] Table 2 Experimental Materials Example 1: Establishment and Validation of a Multi-Indicator NIRS Model for Raw Material Drug Processing—Phellodendron amurense 1.1 Establishment of NIRS models for different batches of Phellodendron amurense 1.1.1 Preparation of different batches of Phellodendron amurense All 10 batches of Phellodendron amurense came from Shaanxi Momede Qi Xue He Pharmaceutical Co., Ltd. After being pulverized, they were passed through a No. 5 sieve (80 mesh, 180 µm ± 7.6 µm) to obtain different batches of Phellodendron amurense samples.

[0037] 1.1.2 Chromatographic conditions An Agilent 5 TC-C18 column (4.5 × 250 mm, 5 µm) was used. The mobile phase consisted of acetonitrile (A) and an aqueous solution containing 0.1% phosphoric acid and 0.1 g / 100 mL sodium dodecyl sulfate (B), with 50:50 isocratic elution, a flow rate of 1.0 mL / min, a column temperature of 30 °C, an injection volume of 10 µL, and a detection wavelength of 345 nm.

[0038] 1.1.3 Preparation of reference solution Accurately weigh palmatine hydrochloride and berberine hydrochloride reference standards, and prepare standard solutions with concentrations of 0.01019 mg / mL and 0.00592 mg / mL respectively using methanol as solvent. Sonicate for 30 min to mix them evenly and prepare a mixed reference solution.

[0039] 1.1.4 Preparation of test solution Accurately weigh approximately 0.1 g of *Phellodendron amurense* sample and transfer it to a 25 mL volumetric flask. Add 20 mL of a 1% hydrochloric acid-methanol solution and extract by sonication for 30 min. After cooling the solution to room temperature, dilute to the mark with methanol, mix thoroughly, and filter. Collect the filtrate and filter it a second time through a 0.22 µm microporous membrane to obtain the *Phellodendron amurense* test solution.

[0040] 1.1.5 Methodological Examination (1) Specificity test Two chemical components to be tested, berberine hydrochloride and berberine hydrochloride, were selected from Phellodendron bark. Blank solvent, mixed reference standard and test sample were injected according to the above chromatographic method.

[0041] (2) Examination of linear relationships Accurately pipette 2, 4, 6, 8, 10, 12, 14, 16, 18, and 20 µL of a mixed reference solution prepared with palmatine hydrochloride and berberine hydrochloride, respectively, and continuously inject the solutions under the chromatographic conditions described above. Plot a standard curve with the concentration of the reference solution (mg) as the abscissa (X) and the peak area integral value as the ordinate (Y), and calculate the regression equation.

[0042] (3) Precision, repeatability, stability and recovery rate were investigated. Six consecutive injections of the mixed solution of Phellodendron amurense reference standard were performed, and the peak area of ​​the samples was measured. Six batches of the same batch of Phellodendron amurense test sample were prepared, and the peak area of ​​the samples was measured. The Phellodendron amurense test sample solution was taken and placed at room temperature for 24 h, and the peak area of ​​the samples was measured at different time periods.

[0043] 1.1.6 Determination of the content of two components in different batches of Phellodendron bark The content of 10 different batches of Phellodendron amurense samples (100 test solutions) was determined. The content of palmatine hydrochloride and berberine hydrochloride in the test solutions was determined by a two-sample parallel determination method, in which each sample was analyzed twice independently, and the average value of the two determinations was taken as the quantitative analysis result of HPLC.

[0044] 1.1.7 NIRS Characteristic Spectrum Acquisition Results of Different Batches of Phellodendron amurense The NIRS instrument was powered on and warmed up for 30 minutes. An integrating sphere diffuse reflectance solid powder averaging workflow was used. Each sample was scanned three times, and the near-infrared spectral scan was performed 32 times per scan. The spectral scanning range was 10,000–4,000 cm⁻¹. -1 The resolution is 8cm. -1 NIRS spectra of 100 samples from 10 different batches of Phellodendron amurense were collected.

[0045] 1.1.8 Division of the calibration set and validation set The Mahalanobis distance of the samples was calculated, and samples exhibiting spectral anomalies were eliminated by analyzing the residuals between HPLC measurements and predicted values. Using stratified random sampling in MATLAB R2021 b software, a calibration set of 80 samples and a validation set of 20 samples were obtained.

[0046] 1.1.9 Spectral Preprocessing Methods When establishing a near-infrared spectral density (NIRS) quantitative analysis model, the integrating sphere mode is selected for solid samples. For berberine hydrochloride and palmatine hydrochloride in *Phellodendron amurense*, preprocessing of the raw spectra is necessary to establish a reliable quantitative analysis model. Common methods include Standard Normal Variable Transform (SNV), which calculates the mean of each spectrum to reduce spectral differences caused by sample particle size variations; Multivariate Scatter Correction (MSC), which uses the average spectrum as a standard and corrects the measured spectrum using a linear regression equation to eliminate spectral scattering effects; and derivative and smoothing techniques to eliminate instrument noise and baseline drift, improving spectral resolution. The calibration set determination coefficient (Rc), validation set determination coefficient (Rp), root mean square error (RMSEC) related to the calibration set, and root mean square error (RMSEP) related to the validation set are used as evaluation indicators of model performance. The smaller the RMSEC value of the calibration set or the RMSEP value of the prediction set, the smaller the model's prediction error. The closer Rc and Rp are to 1, the better the correlation between the near-infrared predicted values ​​and the HPLC measured values. If the above spectral preprocessing methods are not significantly different, a preprocessing method with fewer principal factors can be considered, which can reduce the complexity of the model and improve its sensitivity.

[0047] 1.1.10 Selection of Spectral Modeling Range TQ Analyst 9.0 software systematically calculated the variance of each wavelength and, combined with near-infrared spectral characteristics and related content determination results, identified the optimal modeling band range. Detailed analysis of variance variations between spectral bands successfully reduced multicollinearity in the data and effectively eliminated interference from redundant information, significantly improving the accuracy of model predictions.

[0048] 1.1.11 Selection of the Number of Principal Factors The selection of the number of principal components is crucial in the NIRS modeling process. Too many latent variables can lead to the loss of spectral information and the model's loss of predictive ability (underfitting), while too few latent variables can cause noise pollution and excessive model fluctuations (overfitting). Combining the inflection point method of the (root mean square cross-validation error, RMSECV) curve, the RMSECV first decreases and then tends to plateau as the number of principal components increases. The optimal number of principal components determined by this is used as the number of latent variables.

[0049] 1.1.12 Establishment and Evaluation of Near-Infrared Model A PLS model was established based on the HPLC content determination results of different batches of Phellodendron amurense, as well as the HPLC content determination results of moisture and color.

[0050] 1.1.13 Examination of the validation set results of the near-infrared quantitative model Leave-one-out cross-validation (LOO-CV) was used to evaluate the NIRS model of the content of two chemical components, moisture and color of different batches of Phellodendron amurense. The method involves selecting one sample from m samples in turn each time and using it as the validation set, while the remaining m-1 samples are used as the training set. The average value of the m validation results is used as the final evaluation index.

[0051] 1.2 Methodological Investigation of NIRS Model Establishment for Different Batches of Phellodendron amurense 1.2.1 Specificity test The results showed that the HPLC specificity of Phellodendron bark was good, and there was no interference from the blank solvent. The chromatographic peaks corresponded well. Figure 1 .

[0052] 1.2.2 Examination of Linear Relationships The correlation coefficient R>0.9999, and the linear range meets the requirements for content determination, as shown in Table 3.

[0053] Table 3. Linear relationship results of two components of Phellodendron bark. 1.2.3 Precision, repeatability, stability, and recovery rate were investigated. Six consecutive injections of the *Phellodendron amurense* reference standard mixed solution yielded a peak area RSD < 2% (Table 4), indicating good precision. Six batches of the same batch of *Phellodendron amurense* sample were prepared using the same preparation method, and the peak area RSD was < 4% (Table 5), indicating good repeatability. The *Phellodendron amurense* sample solution was incubated at room temperature for 24 h, and samples were measured at different time points, yielding a peak area RSD < 1% (Table 6), indicating good stability of the sample preparation method. Reference standards, palmatine hydrochloride and berberine hydrochloride stock solutions at concentrations of 6.3542 mg / mL and 7.8513 mg / mL, respectively, were added to 0.05 g of *Phellodendron amurense* sample, with added volumes of 47 µL and 103 µL, respectively. Six batches were prepared using the same preparation method, and the average recovery rate was 95–105%, with an RSD < 1% (Table 7), indicating good recovery.

[0054] Table 4. Precision test results (n=6) Table 5. Results of repeatability testing (n=6) Table 6. Stability test results (n=8) Table 7 Results of the recovery test (n=6) 1.3 NIRS Characteristic Spectrum Acquisition Results of Different Batches of Phellodendron amurense Its spectral features are distinct, and the similarity between the spectra is high. Figure 2 .

[0055] 1.3.1 Division of calibration set and validation set The content, moisture content, and color value of the two target components in each batch of Phellodendron amurense in the validation set were uniformly distributed within the sample. For details of the specific division scheme, please refer to Table 8.

[0056] Table 8. Distribution of calibration and validation sets for moisture, color, and content of two components in Phellodendron amurense. 1.3.2 Spectral Preprocessing Methods Table 9 shows a comparison of the effects of different pretreatment methods. The optimal spectral pretreatment methods selected through comprehensive screening are: berberine hydrochloride (Constant+2stD+SG), berberine hydrochloride (Constant+1stD+SG), Phellodendron amurense moisture (Constant+2stD+ND), and Phellodendron amurense color (Constant+1stD+SG).

[0057] Table 9. Preprocessing results of moisture, color, and two components of different batches of Phellodendron amurense. 1.3.3 Selection of Spectral Modeling Range The results showed that the optimal modeling range for palmatine hydrochloride and berberine hydrochloride was determined to be 7037.62–4196.34 cm⁻¹. - ¹This provides a solid foundation for further research.

[0058] 1.3.4 Selection of the Number of Principal Factors The optimal principal factor numbers for palmatine hydrochloride, berberine hydrochloride, moisture, and color were 4, 3, 10, and 1, respectively, with corresponding RMSECV values ​​of 0.0378, 0.1020, 0.1210, and 0.1990. All indicator models were in a relatively ideal stable state. (See...) Figure 3 .

[0059] 1.3.5 Establishment and Evaluation of Near-Infrared Model A PLS model was established based on the HPLC content determination results of different batches of Phellodendron amurense, as well as the HPLC content determination results of moisture and color. The relationship between the model's predicted values ​​and the reference values, and the residual distribution are shown below. Figure 4As shown, the Rc values ​​for the calibration set of four indicators—palmatine hydrochloride, berberine hydrochloride, moisture, and color—are 0.8607, 0.9556, 0.9696, and 0.9119, respectively, with RMSEC values ​​of 0.0325, 0.0930, 0.0822, and 0.1820. The Rp values ​​for the prediction set are 0.8936, 0.9616, 0.9475, and 0.9713, respectively, with RMSEP values ​​of 0.0303, 0.1000, 0.1070, and 0.1430. All R values ​​are very close to 1, and both RMSEC and RMSEP are very close to 1, indicating that the model has high prediction accuracy and robustness.

[0060] 1.3.6 Examination of the validation set results of the near-infrared quantitative model LOO-CV was performed on two chemical components, moisture content, and colorimetric indices in different batches of *Phellodendron amurense* samples. The results showed that the correlation coefficients of the predicted values ​​for palmatine hydrochloride, berberine hydrochloride, moisture content, and colorimetric values ​​were all greater than 0.9, and the cross-validation error was close to 0, indicating that the near-infrared quantitative model has good generalization ability. (See...) Figure 5 The average relative errors between the predicted and reference values ​​of the two components, moisture content, and colorimetric indexes of 20 batches of Phellodendron amurense from different batches were 4.25%, 0.57%, 1.13%, and 0.19%, respectively. The relative errors were less than 5%, indicating that the model has high prediction accuracy. (See...) Figure 6 .

[0061] Example 2: Raw Material Stage—Establishment and Validation of NIRS Models for Salvia miltiorrhiza from Different Origins 2.1 Establishment of NIRS models for Salvia miltiorrhiza from different origins 2.1.1 Preparation of Salvia miltiorrhiza from different origins Different batches of Salvia miltiorrhiza samples were obtained by pulverizing Salvia miltiorrhiza from different origins and passing it through a No. 5 sieve (80 mesh, 180µm ± 7.6 µm) for NIRS scanning and HPLC content determination.

[0062] 2.1.2 Chromatographic conditions The column temperature was 25℃, the detection wavelength was 286 nm (UV), and the rest was the same as the method under "1.1.2". Gradient elution is shown in Table 10.

[0063] Table 10 Gradient elution table for determining the content of Salvia miltiorrhiza from different origins 2.1.3 Preparation of reference solution Accurately weigh rosmarinic acid, salvianolic acid B, cryptotanshinone, tanshinone I, and tanshinone IIA reference standards, and prepare mixed reference standard solutions with concentrations of 0.01299 mg / mL, 0.08667 mg / mL, 0.00593 mg / mL, 0.00397 mg / mL, and 0.00952 mg / mL using methanol as the solvent.

[0064] 2.1.4 Preparation of test solution Accurately weigh 0.3 g of Danshen sample, select 80% methanol solution as solvent, and prepare it using the same method as in section "1.1.4".

[0065] 2.1.5 Methodological Examination (1) Specificity test The specificity of five characteristic chemical components in Salvia miltiorrhiza—rosmarinic acid, salvianolic acid B, cryptotanshinone, tanshinone I, and tanshinone IIA—was investigated under chromatographic conditions, using blank solvent, mixed reference standard, and test sample solution.

[0066] (2) Examination of linear relationships Accurately pipette 2, 4, 6, 8, 10, 12, 14, 16, 18, and 20 µL of a mixed reference solution prepared with palmatine hydrochloride and berberine hydrochloride, respectively, and continuously inject the solutions under the chromatographic conditions described above. Plot a standard curve with the concentration of the reference solution (mg) as the abscissa (X) and the peak area integral value as the ordinate (Y), and calculate the regression equation.

[0067] (3) Precision, repeatability, stability and recovery rate were investigated. Six consecutive injections of the mixed solution of Danshen reference standard were performed, and the peak area of ​​the samples was measured. Six batches of Danshen test samples of the same batch were prepared, and the peak area of ​​the samples was measured. The Danshen test sample solution was taken and placed at room temperature for 24 h, and the peak area of ​​the samples was measured at different time periods.

[0068] 2.1.6 Determination of the content of five components in Salvia miltiorrhiza from different origins The content of 90 Salvia miltiorrhiza samples (a total of 90 test solutions) from 9 different origins was determined using a two-sample, two-needle assay. The contents of rosmarinic acid, salvianolic acid B, cryptotanshinone, tanshinone I, and tanshinone IIA were determined separately. A two-sample, parallel assay method was used, with each sample being analyzed twice independently. The arithmetic mean of the two measurements was taken as the quantitative analysis result by high-performance liquid chromatography (HPLC).

[0069] 2.1.7 NIRS Characteristic Spectrum Acquisition Results of Salvia miltiorrhiza from Different Origins The NIRS instrument was powered on and warmed up for 30 minutes. An integrating sphere diffuse reflectance solid powder averaging workflow was used. Each sample was scanned three times, and the near-infrared spectral scan was performed 32 times per scan. The spectral scanning range was 10,000–4,000 cm⁻¹. -1 The resolution is 8cm. -1 NIRS spectra of 90 batches of Salvia miltiorrhiza from different origins were collected.

[0070] 2.1.8 Partitioning of the calibration set and validation set Twenty validation set samples and seventy calibration set samples were randomly selected using MATLAB R 2021 b software.

[0071] 2.1.9 Spectral Preprocessing Methods After considering the same spectral preprocessing methods as in "1.1.9", the optimal spectral preprocessing methods were selected as follows: rosmarinic acid SNV+2stD+ND, salvianolic acid BSNV+2stD+ND, cryptotanshinone MSC+1stD+SG, tanshinone IMSC+2stD+SG, tanshinone IIAMSC+1stD+ND, tanshinone moisture SNV, and tanshinone color SNV.

[0072] 2.1.10 Selection of Spectral Modeling Range The interval selection method is the same as "1.1.10".

[0073] 2.1.11 Selection of the Number of Principal Factors The method for selecting the number of principal factors is the same as in "1.1.11".

[0074] 2.1.12 Establishment and Evaluation of Near-Infrared Model A PLS model was established based on the HPLC content of different batches of Salvia miltiorrhiza, as well as the HPLC content determination results of moisture and color.

[0075] 2.1.13 Examination of the validation set results of the near-infrared quantitative model The NIRS model of Salvia miltiorrhiza from different origins was tested using the LOO-CV algorithm.

[0076] 2.2 Methodological Investigation of NIRS Models for Salvia miltiorrhiza from Different Origins 2.2.1 Specificity Test The results showed that the five components were well separated, and the retention times of the peaks of the test sample and the mixed reference standard were consistent. No interfering peaks appeared in the blank solvent, indicating good specificity. Figure 7 .

[0077] 2.2.2 Examination of Linear Relationships The results show that R>0.9999 for all target components, and the linear range meets the requirements for quantitative analysis. Detailed data are shown in Table 11.

[0078] Table 11. Linear Relationship Results of Salvia miltiorrhiza Components 2.2.3 Precision, repeatability, stability, and recovery rate were investigated. Six consecutive injections of the danshen reference standard mixed solution yielded a peak area RSD < 1% (Table 12), indicating good precision. Six batches of the same batch of danshen test sample were prepared using the same test sample preparation method, and the peak area RSD was < 4% (Table 13), indicating good repeatability. The danshen test sample solution was incubated at room temperature for 24 h, and samples were measured at different time points, yielding a peak area RSD < 4% (Table 14), demonstrating the good stability of the test sample preparation method. Reference standards were added to 0.15 g of Salvia miltiorrhiza sample. The concentrations of the stock solutions of rosmarinic acid, salvianolic acid B, cryptotanshinone, tanshinone I, and tanshinone IIA reference standards were 5.3412 mg / mL, 15.5672 mg / mL, 6.8511 mg / mL, 7.3721 mg / mL, and 8.5311 mg / mL, respectively. The added volumes were 71 µL, 380 µL, 32 µL, 10 µL, and 42 µL, respectively. Six batches were prepared according to the test sample preparation method. The average recovery rate was 95-105%, and the RSD was <4%, indicating that the recovery rate was good (Table 15).

[0079] Table 12 Precision test results of five components of Salvia miltiorrhiza (n=6) Table 13. Results of repeatability study of five components of Salvia miltiorrhiza (n=6) Table 14. Stability test results of five components of Salvia miltiorrhiza (n=8) Table 15 Results of the recovery rate test of five components of Salvia miltiorrhiza (n=6) 2.3 NIRS Characteristic Spectrum Acquisition Results of Salvia miltiorrhiza from Different Origins NIRS spectra of 90 batches of Salvia miltiorrhiza from different origins were collected. The spectral characteristics were distinct, and the similarity between the spectra was high. (See attached image.) Figure 8 .

[0080] 2.3.1 Division of calibration set and validation set The content, moisture, and color values ​​of five components of Salvia miltiorrhiza from different origins in the validation set were all evenly distributed except for salvianolic acid B, which had a slightly poorer stability and a larger standard deviation. The data division results are shown in Table 16.

[0081] Table 16 Distribution of calibration and validation sets for moisture, color, and content of five components in Salvia miltiorrhiza. 2.3.2 Spectral Preprocessing Methods The optimal spectral preprocessing methods were selected as follows: rosmarinic acid SNV+2stD+ND, salvianolic acid B SNV+2stD+ND, cryptotanshinone MSC+1stD+SG, tanshinone I MSC+2stD+SG, tanshinone IIA MSC+1stD+ND, tanshinone moisture SNV, and tanshinone color SNV. A comparison of the optimization effects of different preprocessing methods is shown in Table 17.

[0082] Table 17 Results of spectral preprocessing of moisture, color, and five components of Salvia miltiorrhiza from different origins. 2.3.3 Selection of Spectral Modeling Range TQ Analyst 9.0 software recommends the modeling bands for rosmarinic acid, salvianolic acid B, cryptotanshinone, tanshinone I, tanshinone IIA, and the moisture and color of tanshinone as 7339.74~4196.34 cm⁻¹. -1 .

[0083] 2.3.4 Selection of the Number of Principal Factors The optimal principal component numbers for rosmarinic acid, salvianolic acid B, cryptotanshinone, tanshinone I, tanshinone IIA, and the moisture and color of tanshinone were 10, 10, 10, 7, 10, 3, and 7, respectively, with corresponding RMSECV values ​​of 0.2410, 3.1500, 0.2400, 0.0414, 0.2740, 1.1800, and 1.4500. (See...) Figure 9 .

[0084] 2.3.5 Establishment and Evaluation of Near-Infrared Model The Rc values ​​for the quantitative analysis models of the five components, moisture, and color were 0.9904, 0.9914, 0.9895, 0.9505, 9866, 0.7512, and 0.9552, respectively; the RMSEC values ​​were 0.1500, 1.8300, 0.1250, 0.0220, 0.1320, 1.1100, and 1.0600, respectively; the Rp values ​​were 0.9645, 0.9710, 0.9827, 0.8890, 0.9945, 0.6533, and 0.9469, respectively; and the RMSEP values ​​were 0.2870, 3.8300, 0.1660, 0.0402, 0.0852, 1.4500, and 1.0600, respectively. The R-value is close to 1, and RMSEC and RMSEP are close to 0, indicating that the model has good accuracy and robustness. See Figure 10 .

[0085] 2.3.6 Examination of the validation set results of the near-infrared quantitative model The LOO-CV algorithm was used to test the NIRS models for Salvia miltiorrhiza from different origins. The correlation coefficients for rosmarinic acid, salvianolic acid B, cryptotanshinone, tanshinone I, tanshinone IIA, moisture content, and color of Salvia miltiorrhiza were >0.9, and the RMSEV values ​​were all close to 0. The RMSEV value of salvianolic acid B was relatively large, indicating slightly insufficient model generalization ability. (See...) Figure 11 The average relative errors between the predicted and actual values ​​of the five components (moisture, color, and chromaticity) in the 20 batch validation sets were 5.16, 7.85, 6.62, 9.08, 7.82, 6.96, and 1.73, respectively. The average relative error was less than 10%, indicating good predictive accuracy of the model. (See...) Figure 12 .

[0086] Example 3: Establishment and Validation of the NIRS Model for the Production Process—Phellodendron amurense Extract Concentrate 3.1 Establishment of the NIRS model for the concentrated extract of Phellodendron amurense 3.1.1 Preparation of Aqueous Extract of Phellodendron bark This embodiment refers to the extraction method of Shuangshitonglin capsules, employing multiple dynamic reflux extractions of key components from *Phellodendron amurense*. 100 g of *Phellodendron amurense* was weighed and placed in a three-necked flask. A first continuous reflux extraction was performed with 1200 mL of pure water for 90 min. Subsequently, 5 mL samples were taken every 2 min, and an equal amount of solvent was added to maintain the solvent volume. For the second and third extractions, 1000 mL of pure water was added, and dynamic reflux extraction was performed for 60 min. 5 mL samples were taken every 3 min, and an equal amount of solvent was added to maintain the solvent volume. NIRS was used to monitor the sampling points. (See attached image). Figure 13 .

[0087] 3.1.2 Preparation of concentrated Phellodendron bark extract The concentration was carried out according to the concentration method of Phellodendron amurense in the preparation process of Shuangshitonglin capsules. The three aqueous extracts of Phellodendron amurense were combined and mixed, and then heated and concentrated in a beaker. In the early stage of concentration, samples were taken every 5 minutes, and in the later stage of concentration, samples were taken every 3 minutes, with a sample volume of about 5 mL.

[0088] 3.1.3 Chromatographic conditions Same as the method under "1.1.2".

[0089] 3.1.4 Preparation of reference solution Accurately weigh the reference standards of palmatine hydrochloride and berberine hydrochloride, and prepare mixed reference standard solutions with concentrations of 0.01019 mg / mL and 0.00592 mg / mL using methanol as the solvent.

[0090] 3.1.5 Preparation of test solution Accurately measure 1 mL of the water extract-concentrate of Phellodendron bark, and prepare it using methanol with 1% hydrochloric acid as the solvent, and follow the same method as in section "1.1.4".

[0091] 3.1.6 Determination of Chemical Components in Aqueous Extract and Concentrate of Phellodendron bark Aqueous extracts and concentrates of Phellodendron amurense were collected and tested according to the test sample preparation method, resulting in a total of 85 aqueous extract test samples and 80 concentrate test samples. The contents of palmatine hydrochloride and berberine hydrochloride in each test sample were determined sequentially according to the chromatographic conditions. The two-sample parallel determination method was adopted, and each sample was tested twice independently. The arithmetic mean of the two determination data was finally used as the quantitative analysis result of high performance liquid chromatography.

[0092] 3.1.7 Acquisition of NIRS characteristic spectra of Phellodendron amurense aqueous extract-concentrate The sample of concentrated aqueous extract of Phellodendron bark was placed in a 50 mm diameter cuvette and collected using the intelligent transmission liquid averaging working flow mode.

[0093] 3.1.8 Division of calibration set and validation set Using MATLABR 2021b software, 20 and 15 spectral samples were extracted as validation sets, and the remaining 65 and 4 spectral samples were used as calibration sets. The validation sets were uniformly distributed from low to high concentrations.

[0094] 3.1.9 Spectral Pretreatment Method for Aqueous Extract and Concentrate of Phellodendron amurense Taking into account the same criteria as in "1.1.9", the optimal spectral pretreatment methods were selected as follows: water extract of Phellodendron amurense - berberine hydrochloride, water extract of Phellodendron amurense - palmatine hydrochloride MSC+2stD+ND; concentrated extract of Phellodendron amurense - berberine hydrochloride, and concentrated extract of Phellodendron amurense - palmatine hydrochloride Constant+2stD+ND.

[0095] 3.1.10 Selection of Spectral Modeling Interval The optimal modeling range and preferred band range for berberine hydrochloride were obtained through analysis using TQ Analyst 9.0 software.

[0096] 3.1.11 Selection of the Number of Principal Factors The method for selecting the number of principal factors is the same as in "1.1.11".

[0097] 3.1.12 Establishment and Evaluation of Near-Infrared Model Based on HPLC measurements, the optimal wavelength range was selected to establish the PLS model.

[0098] 3.1.13 Examination of the validation set results of the near-infrared quantitative model The LOO-CV algorithm was used to optimize the NIRS model with low content components, and the average error was taken as the model performance index.

[0099] 3.2 Methodological Investigation of the NIRS Model for Phellodendron amurense Extract Concentrate 3.2.1 Specificity Test The water extract and concentrate of Phellodendron amurense showed good specificity for berberine hydrochloride and palmatine hydrochloride, with no interference from the blank solvent. See the relevant chromatograms for details. Figure 14 .

[0100] 3.2.2 Linear Relationship The results showed that palmatine hydrochloride and berberine hydrochloride had a good linear relationship (R>0.999), and the regression equation is shown in Table 18.

[0101] Table 18 Results of linearity study of two components in the aqueous extract-concentrate of Phellodendron amurense. 3.2.3 Precision, repeatability, stability, and recovery rate were investigated. Six consecutive injections of the mixed solution of Phellodendron amurense aqueous extract-concentrate reference solution yielded a peak area RSD < 2% (Table 19), indicating good precision. Six batches of the same batch of Phellodendron amurense aqueous extract-concentrate test sample were prepared using the same test sample preparation method, and the peak area RSD was < 1% (Table 20), indicating good repeatability. The Phellodendron amurense aqueous extract-concentrate test sample solution was incubated at room temperature for 24 h, and samples were measured at different time points, yielding a peak area RSD < 1% (Table 21), indicating good stability of the test sample preparation method. Reference standards were added to 0.5 mL of Phellodendron amurense aqueous extract-concentrate sample. The concentrations of palmatine hydrochloride and berberine hydrochloride mother liquor were 6.3542 mg / mL and 7.8513 mg / mL, respectively. The added volumes were 16 µL and 30 µL, respectively. Six batches were prepared according to the test sample preparation method. The average recovery rate was 95-105%, and the RSD was <1% (Table 22), indicating that the recovery rate was good.

[0102] Table 19. Precision test results of two components in the aqueous extract-concentrate of Phellodendron amurense (n=6) Table 20. Repeatability test results of two components in the aqueous extract-concentrate of Phellodendron amurense (n=6) Table 21. Stability test results of two components in the aqueous extract-concentrate of Phellodendron amurense (n=8) Table 22 Results of the spiking recovery rate test of two components in the aqueous extract-concentrate of Phellodendron amurense (n=6) 3.3 Acquisition of NIRS Characteristic Spectra of Phellodendron amurense Aqueous Extract-Concentrate The samples of the concentrated aqueous extract of Phellodendron amurense were placed in 50 mm diameter cuvettes and acquired using the intelligent transmission liquid average working flow mode. NIRS spectra of 85 batches of aqueous extracts and 55 batches of concentrates were obtained, as shown below. Figure 15 .

[0103] 3.3.1 Division of calibration set and validation set Using MATLABR 2021b software, 20 and 15 spectral samples were extracted as validation sets, and the remaining 65 and 4 spectral samples were used as calibration sets. The validation sets were uniformly distributed from low to high concentrations. The data partitioning results are shown in Table 23.

[0104] Table 23 shows the content distribution of two components of Phellodendron amurense per mL of extract-concentrate in the calibration and validation sets. 3.3.2 Spectral Pretreatment Method for Phellodendron amurense Aqueous Extract-Concentrate Taking all factors into consideration, the optimal spectral pretreatment methods were selected as follows: Phellodendron amurense aqueous extract-berberine hydrochloride, Phellodendron amurense aqueous extract-palmatine hydrochloride MSC+2stD+ND; Phellodendron amurense concentrate-berberine hydrochloride, and Phellodendron amurense concentrate-palmatine hydrochloride Constant+2stD+ND, as shown in Table 24.

[0105] Table 24. Spectral preprocessing results of two components in the aqueous extract-concentrate of Phellodendron amurense. 3.3.3 Selection of Spectral Modeling Range Analysis using TQ Analyst 9.0 software determined the optimal modeling range for berberine hydrochloride to be 7173.90–5364.99 cm⁻¹. - ¹, The preferred wavelength range for palmatine hydrochloride is 5322.57–4593.61 cm⁻¹. - ¹.

[0106] 3.3.4 Selection of the Number of Principal Factors The optimal principal factor numbers for water-extracted berberine hydrochloride, water-extracted palmatine hydrochloride, concentrated berberine hydrochloride, and concentrated palmatine hydrochloride were 6, 8, 5, and 5, respectively, with corresponding RMSECVs of 0.0368, 0.0130, 0.1180, and 0.0479. (See...) Figure 16 .

[0107] 3.3.5 Establishment and Evaluation of Near-Infrared Model Based on HPLC measurements, an optimal wavelength range was selected to establish a PLS model. The calculated values ​​and reference values ​​are shown below. Figure 17 The Rc values ​​for water-extracted berberine hydrochloride, water-extracted palmatine hydrochloride, concentrated berberine hydrochloride, and concentrated palmatine hydrochloride were 0.9860, 0.9909, 0.9130, and 0.9085, respectively; the RMSEC values ​​were 0.0300, 0.0096, 0.0823, and 0.0336, respectively; the Rp values ​​were 0.9890, 0.9908, 0.9074, and 0.9034, respectively; and the RMSEP values ​​were 0.0306, 0.0110, 0.1220, and 0.0496, respectively. The RMSEC and RMSEP values ​​are close to 1, indicating that this model can accurately predict the content of the samples.

[0108] 3.3.6 Examination of the validation set results of the near-infrared quantitative model LOO-CV results showed that the RMSECV of the two components in the water extract and concentrate was >0.9, and the validation error was close to 0, indicating that the model has good predictive ability. (See...) Figure 18The average relative errors between the predicted and measured values ​​of the 20 batches of sample validation sets were 6.23%, 4.69%, 5.48%, and 6.38%, respectively. All average relative errors were below 7%, indicating good model predictive performance. (See...) Figure 19 .

[0109] Example 4: Establishment and Validation of the NIRS Model for the Production Process—Ethanol Extract Concentrate of Salvia miltiorrhiza 4.1 Establishment of the NIRS model for the concentrated ethanol extract of Salvia miltiorrhiza 4.1.1 Preparation of Tanshinone Alcohol Extract Following the standardized production process of Shuangshitonglin capsules, Danshen raw materials were extracted in steps. 100g of Danshen was weighed and placed in a three-necked flask reactor. Two hot reflux extractions were performed, each lasting 60 min, with 600 mL of 80% ethanol solution added. During the extraction process, 5 mL of sample was collected every 3 min, and an equal volume of 80% ethanol was added to maintain a constant solvent volume. Two batches of samples were prepared in parallel for simultaneous extraction.

[0110] 4.1.2 Preparation of Danshen Concentrate The concentration of danshen was carried out according to the concentration method of danshen in the preparation process of Shuangshitonglin capsules. The ethanol extracts of danshen were combined and mixed, and then heated and concentrated in a rotary evaporator. Samples were taken once every 5 minutes, and the sample volume was 5 mL each time.

[0111] 4.1.3 Chromatographic conditions The column temperature was 25 ℃, the UV detection wavelength was 286 nm, and the rest of the methods were the same as those in section "1.1.2". Under these chromatographic conditions, the resolution of each peak was good, all greater than 1.5. The gradient elution is shown in Table 25.

[0112] Table 25 Gradient elution table for content determination of tanshinone alcohol extract-concentrate. 4.1.4 Preparation of reference solution Accurately weigh rosmarinic acid, salvianolic acid B, cryptotanshinone, tanshinone I, and tanshinone IIA reference standards, and prepare mixed reference standard solutions with concentrations of 0.01299 mg / mL, 0.08667 mg / mL, 0.00593 mg / mL, 0.00397 mg / mL, and 0.00952 mg / mL using methanol as the solvent.

[0113] 4.1.5 Preparation of test solution Take 1 mL of the tanshinone alcohol extract-concentrate, using 80% methanol as the solvent, and follow the same procedure as in section "1.1.4".

[0114] 4.1.6 Determination of Chemical Components in Tanshinone Alcohol Extract-Concentrate Take the tanshinone ethanol extract-concentrate and prepare it according to the test sample preparation method. A total of 80 tanshinone ethanol extract test sample solutions and 80 concentrated solution test samples were prepared. The contents of rosmarinic acid, salvianolic acid B, cryptotanshinone, tanshinone I, and tanshinone IIA in each test sample were determined. The content determination adopted the double-sample parallel determination method, and each sample was detected twice independently. Finally, the arithmetic mean of the two determination data was used as the quantitative analysis result of high performance liquid chromatography.

[0115] 4.1.7 NIRS Characteristic Spectrum Acquisition of Tanshinone Alcohol Extract-Concentrate Collect the data using the same method as described in section “3.1.7”.

[0116] 4.1.8 Partitioning of the calibration set and the validation set The MATLABR 2021b software extracted 20 spectral samples as the validation set and the remaining 60 spectral samples as the calibration set. The validation set was uniformly distributed from low concentration to high concentration.

[0117] 4.1.9 Spectral Preprocessing Methods The same spectral preprocessing method as under item "1.1.9" was selected comprehensively.

[0118] 4.1.10 Selection of Spectral Modeling Interval The optimal modeling range and preferred band range were obtained through analysis using TQ Analyst 9.0 software.

[0119] 4.1.11 Selection of the Number of Principal Factors The method for selecting the number of principal factors is the same as in "1.1.11".

[0120] 4.1.12 Establishment and Evaluation of Near-Infrared Model Based on HPLC measurements, the optimal wavelength range was selected to establish the PLS model.

[0121] 4.1.13 Examination of the validation set results of the near-infrared quantitative model The LOO-CV algorithm was used to optimize the NIRS model with low content components, and the average error was taken as the model performance index.

[0122] 4.2 NIRS Model Methodological Investigation of Danshen Ethanol Extract Concentrate 4.2.1 Specificity Test The five chromatographic peaks of rosmarinic acid, salvianolic acid B, cryptotanshinone, tanshinone I, and tanshinone IIA in the ethanol extract-concentrate of tanshinone showed good resolution, and no interfering peaks were observed in the blank solvent. Figure 20 .

[0123] 4.2.2 Examination of Linear Relationships The linear relationship is good (R>0.999), and the regression equation is shown in Table 26.

[0124] Table 26. Linearity Study Results of Two Components in Tanshinone Alcohol Extract-Concentrate 4.2.3 Precision, repeatability, stability, and recovery rate were investigated. Six consecutive injections of the tanshinone ethanol extract-concentrate reference solution yielded a peak area RSD < 2% (Table 27), indicating good precision. Six batches of the same batch of tanshinone ethanol extract-concentrate were prepared using the same preparation method, and the peak area RSD was < 4% (Table 28), indicating good repeatability. The tanshinone ethanol extract-concentrate test solution was incubated at room temperature for 24 hours, and samples were measured at different time points, yielding a peak area RSD < 2% (Table 29), demonstrating the good stability of the test sample preparation method. Reference standards were added to 0.5 mL of tanshinone ethanol extract-concentrate sample. The concentrations of the mother liquor of rosmarinic acid, salvianolic acid B, cryptotanshinone, tanshinone I, and tanshinone IIA reference standards were 5.3412 mg / mL, 15.5672 mg / mL, 6.8511 mg / mL, 7.3721 mg / mL, and 8.5311 mg / mL, respectively. The added volumes were 40 µL, 329 µL, 18 µL, 13 µL, and 18 µL. Six batches were prepared according to the test sample preparation method. The average recovery rate was 95-105%, and the RSD was <2% (Table 30), indicating that the recovery rate was good.

[0125] Table 27 Precision test results (n=6) Table 28 Results of Repeatability Tests (n=6) Table 29 Stability test results (n=8) Table 30 Results of the sample recovery test (n=6) 4.3 NIRS Characteristic Spectrum Acquisition of Tanshinone Alcohol Extract-Concentrate Common characteristic spectra were collected from the near-infrared raw spectra of 85 batches of tanshinone ethanol extract and 79 batches of tanshinone concentrate. Due to the increased concentration, some spectra of the tanshinone ethanol concentrate showed NIRS overload, while the remaining spectra were good. (See attached image) Figure 21 .

[0126] 4.3.1 Partitioning of the calibration set and the validation set The MATLABR 2021b software extracted 20 spectral samples as the validation set and the remaining 60 spectral samples as the calibration set. The validation set was uniformly distributed from low concentration to high concentration. The data partitioning results are shown in Table 31.

[0127] Table 31. Content distribution of two components of Salvia miltiorrhiza per mL of extract-concentrate in the calibration and validation sets. 4.3.2 Spectral Preprocessing Methods Based on comprehensive selection, the optimal spectral preprocessing method for alcohol-extracted rosmarinic acid, alcohol-extracted salvianolic acid B, alcohol-extracted cryptotanshinone, alcohol-extracted tanshinone I, and alcohol-extracted tanshinone IIA is MSC+2stD+ND; for concentrated salvianolic acid B, concentrated tanshinone IIA is MSC+1stD+ND, and for concentrated rosmarinic acid, concentrated cryptotanshinone, and concentrated tanshinone I, it is SNV. Different spectral preprocessing methods are shown in Table 32.

[0128] Table 32. Spectral preprocessing results of five components in the tanshinone alcohol extract-concentrate. 4.3.3 Selection of Spectral Modeling Range The wavelength range of rosmarinic acid is 7173.90–5364.99 cm⁻¹. -1 The modeling band for salvianolic acid B is 5322.57~4593.61cm. -1 The modeling band for cryptotanshinone is 5322.57~4593.61 cm⁻¹. -1 The modeling band for tanshinone I is 5322.57~4593.61cm. -1 The modeling band for tanshinone IIA is 5322.57–4593.61 cm⁻¹. -1 .

[0129] 4.3.4 Selection of the Number of Principal Factors The optimal principal component numbers for ethanol-extracted rosmarinic acid, ethanol-extracted salvianolic acid B, ethanol-extracted cryptotanshinone, ethanol-extracted tanshinone I, ethanol-extracted tanshinone IIA, concentrated rosmarinic acid, concentrated salvianolic acid B, concentrated cryptotanshinone, concentrated tanshinone I, and concentrated tanshinone IIA were 10, 10, 9, 9, 9, 8, 5, 2, 4, and 9, respectively, with RMSECV values ​​of 0.0218, 0.7890, 0.333, 0.0365, 0.0432, 0.1780, 3.4600, 0.0272, 0.0378, and 0.0433, respectively. (See...) Figure 22 .

[0130] 4.3.5 Establishment and Evaluation of Near-Infrared Model The Rc values ​​for the five component quantitative analysis models were 0.9860, 0.9909, 0.9130, and 0.9085, respectively; the RMSEC values ​​were 0.0300, 0.0096, 0.0823, and 0.0336, respectively; the Rp values ​​were 0.9890, 0.9908, 0.9074, and 0.9034, respectively; and the RMSEP values ​​were 0.0306, 0.0110, 0.1220, and 0.0496, respectively. The R values ​​are close to 1, and the RMSEC and RMSEP values ​​are close to 0, indicating that the model exhibits good accuracy and robustness. (See...) Figure 23 .

[0131] 4.3.6 Examination of the validation set results of the near-infrared quantitative model The LOO-CV algorithm was used to analyze five components of the tanshinone alcohol extract-concentrate. The correlation coefficients of alcohol-rosmarinic acid, alcohol-tanshinone B, alcohol-cryptotanshinone, alcohol-tanshinone I, alcohol-tanshinone IIA, concentrate-rosmarinic acid, concentrate-tanshinone B, concentrate-cryptotanshinone, concentrate-tanshinone I, and concentrate-tanshinone IIA were all greater than 0.9, and the cross-validation error was close to 0. Figure 24 The average relative errors between the predicted and actual values ​​of the five components in the validation set of 20 batches of Tanshinone ethanol extract-concentrate were 5.44%, 5.28%, 4.02%, 3.70%, 4.73%, 5.84%, 3.55%, 5.11%, 5.82%, and 4.75%, respectively. All average relative errors were less than 6%, indicating good predictive accuracy of the model. Figure 25 .

[0132] Example 5: Production Process—Establishment and Validation of NIRS Model for Six Types of Traditional Chinese Medicine Extracts and Concentrates 5.1 Establishment of NIRS Models for Six Types of Traditional Chinese Medicine Extracts and Concentrates 5.1.1 Preparation of a mixed aqueous extract of six traditional Chinese medicines 32 g of Piper longum, 32 g of Patrinia scabiosaefolia, 32 g of Plantago asiatica (wrapped and decocted), 19 g of Talcum (wrapped and decocted), 16 g of Poria cocos, and 32 g of Salvia miltiorrhiza were placed in a three-necked flask, and 1956 mL of purified water was added. The mixture was refluxed for 90 min, followed by another 1630 mL of purified water and refluxed for 90 min. The interval between the two extractions was 6 min. A total of 80 samples were prepared, each approximately 5 mL. To ensure sufficient extraction volume, purified water was added after each extraction, maintaining a constant quantity and volume.

[0133] 5.1.2 Preparation of a concentrated solution of six kinds of traditional Chinese medicine The mixed aqueous extract was concentrated by combining the liquids from the two extractions and transferring them to a rotary evaporator for heating and concentration. Samples were collected every 5 minutes, with each sample being approximately 5 mL.

[0134] 5.1.3 Chromatographic conditions The UV detection wavelength was 230 nm, and the other conditions were the same as in "1.1.2" gradient elution (see Table 33). Under these chromatographic conditions, the peak resolution of each group was greater than 1.5.

[0135] Table 33 Gradient Elution Table for Determination of Concentrated Content in Mixed Aqueous Extracts 5.1.4 Preparation of reference solution Accurately weigh genipin, chlorogenic acid, verbascoside, and salvianolic acid B reference standards, and prepare mixed reference solutions with methanol at concentrations of 0.04875 mg / mL, 0.00439 mg / mL, 0.00446 mg / mL, and 0.07456 mg / mL, respectively.

[0136] 5.1.5 Preparation of test solution Take 1 mL of the mixed water extract-concentrate, using 60% as the solvent, and follow the same procedure as in section "1.1.4".

[0137] 5.1.6 Determination of Chemical Components in Mixed Water Extract-Concentrate Take the mixed water extract-concentrate and prepare the test sample according to the test sample preparation method to obtain a total of 80 mixed water extract test sample solutions and 80 concentrate test samples. The contents of genipinic acid, chlorogenic acid, verbascoside, and salvianolic acid B in each test sample were determined sequentially according to the chromatographic conditions. The content determination of the mixed water extract-concentrate was carried out using the double-sample parallel determination method. Each sample was detected twice independently, and the arithmetic mean of the two determination data was finally used as the quantitative analysis result of HPLC.

[0138] 5.1.7 Acquisition of Characteristic Spectra in Near-Infrared Spectra Using the same method as described in section “3.1.8”, a total of 81 batches of mixed water extracts and 80 batches of mixed concentrates were collected for near-infrared raw spectra.

[0139] 5.1.8 Partitioning of the calibration set and the validation set The MATLABR 2021b software randomly selected 20 spectral samples from the mixed aqueous extract as the validation set and the remaining 61 spectral samples as the calibration set. The mixed concentrate also used 20 spectral samples as the validation set and the remaining 60 spectral samples as the calibration set.

[0140] 5.1.9 Spectral Preprocessing Methods The same spectral preprocessing method as under item "1.1.9" was selected comprehensively.

[0141] 5.1.10 Selection of Spectral Modeling Interval The optimal modeling range and preferred band range were obtained through analysis using TQ Analyst 9.0 software.

[0142] 5.1.11 Selection of the Number of Principal Factors The method for selecting the number of principal factors is the same as in "1.1.11".

[0143] 5.1.12 Establishment and Evaluation of Near-Infrared Model Based on HPLC measurements, the optimal wavelength range was selected to establish the PLS model.

[0144] 5.1.13 Examination of the validation set results of the near-infrared quantitative model The LOO-CV algorithm was used to optimize the NIRS model with low content components, and the average error was taken as the model performance index.

[0145] 5.2 Methodological Investigation on the Establishment of NIRS Models for Six Traditional Chinese Medicine Extracts and Concentrates 5.2.1 Specificity Test Genipinic acid, chlorogenic acid, verbascoside, and salvianolic acid B were the analytes in the mixed water extract-concentrate. Blank solvent, mixed reference standards, and the test sample were injected according to the chromatographic conditions. All five chromatographic peaks of the test sample corresponded to those of the reference standards. The blank solvent showed no interference. (See attached image). Figure 26 .

[0146] 5.2.2 Examination of Linear Relationships R>0.998, indicating a good linear relationship. The regression equation is shown in Table 34.

[0147] Table 34 Results of linearity study of four components in the mixed water extract-concentrate 5.2.3 Precision, repeatability, stability, and recovery rate were investigated. Six consecutive injections of the mixed water extract-concentrate reference solution yielded a peak area RSD < 2% (Table 35), indicating good precision. Six batches of the same batch of mixed water extract-concentrate sample were prepared using the same preparation method, and the peak area RSD was < 1% (Table 36), indicating good repeatability. The mixed water extract-concentrate sample solution was incubated at room temperature for 24 h, and samples were measured at different time points, yielding a peak area RSD < 3% (Table 37), indicating good stability of the sample preparation method. Reference standards were added to 0.5 mL of the mixed water extract-concentrate sample. The concentrations of the reference standard mother liquor for genipin, chlorogenic acid, verbascoside, and salvianolic acid B were 0.3152 mg / mL, 0.2112 mg / mL, 0.8257 mg / mL, and 15.5672 mg / mL, respectively. The added volumes were 555 µL, 29 µL, 20 µL, and 188 µL, respectively. Six batches were prepared according to the test sample preparation method. The average recovery rate was 95-105%, and the RSD was <3% (Table 38), indicating that the recovery rate was good.

[0148] Table 35 Precision test results of four components in the mixed water extract-concentrate (n=6) Table 36. Repeatability test results of the four components of the mixed water extract-concentrate (n=6) Table 37. Stability test results of four components in the mixed water extract-concentrate (n=8) Table 38 Results of the spiking recovery test of four components in the mixed water extract-concentrate (n=6) 5.3 Acquisition of characteristic spectra in near-infrared spectroscopy Near-infrared raw spectra of 81 batches of mixed aqueous extracts and 80 batches of mixed concentrates were collected. The average spectrum is shown in the figure. Figure 27 .

[0149] 5.3.1 Partitioning of the calibration set and the validation set MATLAB R 2021b software randomly selected 20 spectral samples from the mixed aqueous extract as the validation set, and the remaining 61 spectral samples as the calibration set. Similarly, 20 spectral samples from the mixed concentrate were used as the validation set, and the remaining 60 spectral samples were used as the calibration set. The validation set was uniformly distributed from low to high concentration, and the data partitioning results are shown in Table 39.

[0150] Table 39. Content distribution of four components per mL of extract-concentrate in the calibration and validation sets of the mixed water extraction-concentration solution. 5.3.2 Spectral Preprocessing Methods The following were selected: mixed water extract-genipin glycoside, mixed water extract-vernaculoside MSC, mixed water extract-chlorogenic acid MSC+2stD+ND, mixed water extract-tanshinone B MSC+1stD+ND; mixed concentrated-genipin glycoside MSC+1stD, mixed concentrated-chlorogenic acid MSC+2stD+SG, mixed concentrated-vernaculoside, mixed concentrated-tanshinone B Constant+1stD, as shown in Table 40.

[0151] Table 40 Results of spectral pretreatment of five components in the mixed water extract-concentrate 5.3.3 Selection of Spectral Modeling Range Using TQ Analyst 9.0 software, the optimal modeling band for genistein, chlorogenic acid, verbascoside, and salvianolic acid B was selected as 7173.90–5364.99 cm⁻¹. -1 .

[0152] 5.3.4 Selection of the Number of Principal Factors The optimal principal component numbers for the mixed water extracts of geniposide, chlorogenic acid, verbascoside, and salvianolic acid B, as well as the mixed concentrated extracts of geniposide, chlorogenic acid, verbascoside, and salvianolic acid B, were 6, 6, 3, 8, 6, 6, 8, and 8, respectively, with RMSECV values ​​of 0.0107, 0.0040, 0.0010, 0.6720, 0.1180, 0.0092, 0.0044, and 0.8760, respectively. (See [reference needed]). Figure 28 .

[0153] 5.3.5 Establishment and Evaluation of Near-Infrared Model The Rc values ​​for the four quantitative analysis models were 0.8289, 0.9414, 0.7692, 0.9592, 0.9085, 0.8339, 0.9565, and 0.9573, respectively; the RMSEC values ​​were 0.0087, 0.0028, 0.0009, 0.5100, 0.1090, 0.0073, 0.0040, and 0.8120, respectively; and the Rp values ​​were 0.9096, 0.9908, 0.7705, 0.6689, 0.9316, 0.9683, 0.8350, and 0.9720, respectively. The R-values ​​are 0.9777, and the RMSEP values ​​are 0.0101, 0.0028, 0.0012, 0.6630, 0.0818, 0.0101, 0.0030, and 0.5950, respectively. The R-value is close to 1, and the RMSEC and RMSEP are close to 0. This model exhibits good accuracy and robustness. (See...) Figure 29 .

[0154] 5.3.6 Examination of the validation set results of the near-infrared quantitative model The NIRS model for the tested components was validated using the LOO-CV algorithm. The correlation coefficients for the mixed water extracts of geniposide, chlorogenic acid, verbascoside, salvianolic acid B, concentrated geniposide, concentrated chlorogenic acid, concentrated verbascoside, and concentrated salvianolic acid B were all greater than 0.9, and the cross-validation error was close to 0. This indicates that the near-infrared quantification model has good predictive ability. Figure 30 The results of 20 batches of samples in the validation set of the mixed water extract-concentrate showed that the average relative errors between the predicted and actual values ​​of each component were 3.14%, 5.59%, 4.67%, 3.79%, 5.53%, 5.16%, 4.50%, and 4.22%, respectively. The average relative error was less than 6%, indicating that the model has good predictive ability. (See...) Figure 31 .

[0155] Example 6 Finished Product Stage—Establishment and Validation of the NIRS Model for Shuangshitonglin Capsules 6.1 Establishment of the NIRS model of Shuangshitonglin capsules 6.1.1 Chromatographic conditions The UV detection wavelength was 254 nm, and the rest of the method was the same as in section "1.1.2". Under these chromatographic conditions, the peak resolution of each group was greater than 1.5, and the gradient elution is shown in Table 41.

[0156] Table 41 Gradient elution table for the determination of Shuangshitonglin capsule content. 6.1.2 Preparation of reference solution Accurately weigh verbascoside, salvianolic acid B, berberine hydrochloride, and tanshinone IIA reference standards, and prepare mixed reference solutions with methanol at concentrations of 0.0143 mg / mL, 0.2539 mg / mL, 0.0729 mg / mL, and 0.02524 mg / mL, respectively.

[0157] 6.1.3 Preparation of test solution Take 2.5 g of the contents of Shuangshitonglin capsules, accurately weigh them, use 70% methanol as solvent, and extract by ultrasonication for 40 min. The rest of the method is the same as in section "1.1.4".

[0158] 6.1.4 Preparation of negative test solution According to the prescription ratio (32 g of Phellodendron bark, 64 g of Dioscorea hypoglauca, 64 g of Patrinia scabiosaefolia, 10 g of Indigo naturalis, 38 g of Talcum, 64 g of Plantago asiatica, 32 g of Acorus tatarinowii, 32 g of Poria cocos, 32 g of Atractylodes lancea, and 64 g of Salvia miltiorrhiza), negative samples lacking Plantago asiatica, Phellodendron bark, and Salvia miltiorrhiza were prepared according to the preparation process of Shuangshi Tonglin Capsules. The negative test solutions were prepared according to the method in "6.1.3".

[0159] 6.1.5 Methodological Examination (1) Specificity test Four characteristic components (verrucoumarin, salvianolic acid B, berberine hydrochloride, and tanshinone IIA) from the contents of Shuangshitonglin capsules were analyzed. Blank solvent, mixed reference standard, test sample, and negative test sample were injected separately according to chromatographic conditions.

[0160] (2) Examination of linear relationships The same method as under “1.1.5 (2)” is used for the examination.

[0161] (3) Precision, repeatability, stability and recovery rate were investigated. Six consecutive injections of the Shuangshitonglin capsule reference solution were performed, and the peak area of ​​the samples was measured. Six batches of the same batch of Shuangshitonglin capsule test sample were prepared, and six batches of the same batch of Shuangshitonglin capsule samples were prepared using the same test sample preparation method. The peak area of ​​the samples was measured. The Shuangshitonglin capsule test sample solution was taken and placed at room temperature for 24 h. The peak area of ​​the samples was measured at different time periods.

[0162] 6.1.6 Determination of Chemical Components in Shuangshitonglin Capsules Ten batches of Shuangshitonglin capsules were taken, and eight batches of test samples were prepared for each batch according to the above-mentioned test sample preparation method, resulting in a total of 80 batches of Shuangshitonglin capsule test sample solutions. The contents of verbascoside, salvianolic acid B, berberine hydrochloride, and tanshinone IIA in the 80 batches of samples were determined sequentially according to the chromatographic conditions. The two-sample parallel determination method was adopted, and each sample was detected twice independently. Finally, the arithmetic mean of the two determination data was used as the quantitative analysis result of high performance liquid chromatography.

[0163] 6.1.7 NIRS Characteristic Spectrum Acquisition of Shuangshitonglin Capsules Using the same method as in section “1.1.7”, near-infrared raw spectra of 80 batches of samples were collected.

[0164] 6.1.8 Partitioning of the calibration set and the validation set The MATLAB R 2021b software randomly selected 20 spectral samples as the validation set, and the remaining 60 spectral samples as the calibration set. 6.1.9 Spectral Preprocessing Methods The same spectral preprocessing method as under item "1.1.9" was selected comprehensively.

[0165] 6.1.10 Selection of Spectral Modeling Interval The optimal modeling range and preferred band range were obtained through analysis using TQ Analyst 9.0 software.

[0166] 6.1.11 Selection of the Number of Principal Factors The method for selecting the number of principal factors is the same as in "1.1.11".

[0167] 6.1.12 Establishment and Evaluation of Near-Infrared Model Based on HPLC measurements, the optimal wavelength range was selected to establish the PLS model.

[0168] 6.1.13 Examination of the validation set results of the near-infrared quantitative model The four components and two physicochemical indicators of Shuangshitonglin capsules were tested using LOO-CV as the indicator, and the average error was taken as the model performance index.

[0169] 6.2 Methodological Investigation on the Establishment of the NIRS Model of Shuangshitonglin Capsules 6.2.1 Specificity Test Four characteristic components (verrucoumarin, salvianolic acid B, berberine hydrochloride, and tanshinone IIA) from the contents of Shuangshitonglin capsules were analyzed. Blank solvent, mixed reference standard, test sample, and negative sample were injected separately according to chromatographic conditions. Results showed good peak resolution for each component, consistent retention times between the test sample and reference standard peaks, and no interfering peaks in the blank solvent and negative sample. This method demonstrates good specificity. (See attached image) Figure 32 .

[0170] 6.2.2 Examination of Linear Relationships R>0.999, indicating a good linear relationship (see Table 42).

[0171] Table 42 Results of linearity analysis of four components in Shuangshitonglin Capsules 6.2.3 Precision, repeatability, stability, and recovery rate were investigated. Six consecutive injections of the Shuangshitonglin capsule reference solution yielded a peak area RSD < 2% (Table 43), indicating good precision. Six batches of the same batch of Shuangshitonglin capsule test sample were prepared using the same test sample preparation method, and the peak area RSD was < 5% (Table 44), indicating good repeatability. The Shuangshitonglin capsule test sample solution was incubated at room temperature for 24 hours, and samples were measured at different time points, yielding a peak area RSD < 2% (Table 45), indicating good stability of the test sample preparation method. Reference standards were added to 1.25 g of Shuangshitonglin capsules. The concentrations of the reference standard stock solutions of verbascoside, salvianolic acid B, berberine hydrochloride, and tanshinone IIA were 0.8257 mg / mL, 15.5672 mg / mL, 7.8513 mg / mL, and 8.5311 mg / mL, respectively. The volumes of the reference standard stock solutions added were 334 µL, 523 µL, 269 µL, and 226 µL, respectively. Six batches were prepared according to the test sample preparation method. The average recovery rate was 95-105%, and the RSD was <3% (Table 46), indicating that the recovery rate was good.

[0172] Table 43 Precision test results of four components in Shuangshitonglin Capsules (n=6) Table 44. Repeatability test results of four components in Shuangshitonglin Capsules (n=6) Table 45. Stability test results of four components in Shuangshitonglin Capsules (n=8) Table 46 Results of the recovery rate test of four components in Shuangshitonglin Capsules (n=6) 6.3 NIRS Characteristic Spectrum Acquisition of Shuangshitonglin Capsules Near-infrared raw spectra of 80 batches of samples were collected, and the average spectra are shown below. Figure 33 .

[0173] 6.3.1 Partitioning of the calibration set and the validation set The MATLABR 2021b software randomly selected 20 spectral samples as the validation set and the remaining 60 spectral samples as the calibration set. The data partitioning results are shown in Table 47.

[0174] Table 47 Content distribution of four components in the contents of Shuangshitonglin capsules in the calibration and validation sets. 6.3.2 Spectral Preprocessing Methods After comprehensive evaluation, verbascoside SNV+1stD, tanshinone B SNV+1stD, berberine hydrochloride MSC+2stD+SG, tanshinone IIA MSC+1stD, moisture content of Shuangshitonglin capsules Constant+1stD+ND, and color content of Shuangshitonglin capsules Constant+1stD were selected as the best spectral preprocessing methods. The different preprocessing methods and results are shown in Table 48.

[0175] Table 48. Spectral preprocessing results of four components of Shuangshitonglin capsules 6.3.3 Selection of Spectral Modeling Range The TQ Analyst 9.0 software was used to select the verbascoside mode band as 7247.18–4180.92 cm⁻¹. -1 The modeling band for salvianolic acid B is 7172.15–4217.59 cm⁻¹. -1 The modeling band for berberine hydrochloride is 7264.71–4131.90 cm⁻¹. -1 The modeling band for tanshinone IIA is 7230.03–4167.23 cm⁻¹. -1 The water modeling wavelength for Shuangshitonglin capsules is 7268.57~4167.73 cm⁻¹. -1 The chromaticity modeling band for Shuangshitonglin capsules is 7268.63~4167.11 cm⁻¹. -1 .

[0176] 6.3.4 Selection of the Number of Principal Factors The optimal principal component numbers for the moisture content and color values ​​of verbascoside, salvianolic acid B, berberine hydrochloride, tanshinone IIA, and Shuangshitonglin capsules were 8, 10, 7, 10, 10, and 10, respectively, with RMSECV values ​​of 0.0219, 0.2990, 0.0794, 0.1030, 0.0876, and 0.6160, respectively. (See...) Figure 34 .

[0177] 6.3.5 Establishment and Evaluation of Near-Infrared Model The Rc values ​​of the four component quantitative analysis models and the two Shuangshitonglin capsule indices were 0.9821, 0.9939, 0.9870, 0.9853, 0.9742, and 0.9753, respectively; the RMSEC values ​​were 0.0132, 0.1240, 0.0249, 0.0368, 0.0593, and 0.2770, respectively; the Rp values ​​were 0.9805, 0.9856, 0.8968, 0.9772, 0.9753, and 0.9061, respectively; and the RMSEP values ​​were 0.0226, 0.1960, 0.0668, 0.0938, 0.0631, and 0.4430, respectively. The R values ​​are close to 1, and the RMSEC and RMSEP values ​​are close to 0, indicating that the model has good accuracy and robustness. (See...) Figure 35 .

[0178] 6.3.6 Examination of the validation set results of the near-infrared quantitative model The LOO-CV was used to test four components and two physicochemical indicators of Shuangshitonglin capsules. The results showed that the correlation coefficients of verbascoside, salvianolic acid B, berberine hydrochloride, tanshinone IIA, moisture content of Shuangshitonglin capsules, and colorimetric value of Shuangshitonglin capsules were all greater than 0.9, and the cross-validation error was close to 0. The near-infrared quantitative model has good predictive ability. Figure 36 In the validation set of Shuangshitonglin capsules, the average relative errors between the predicted and actual values ​​of four chemical components and physicochemical indicators in 20 batches of samples were 6.61%, 2.87%, 4.25%, 4.71%, 0.90%, and 0.93%, respectively. All average relative errors were less than 7%, indicating good predictive accuracy of the model. Figure 37 .

[0179] Example 7 This embodiment provides a detection system, which includes an Antaris II near-infrared spectrometer (ThermoFisher Scientific), an industrial control computer, and supporting data processing software. The functions of each module are as follows: Near-infrared spectroscopy acquisition module: Consists of an Antaris II near-infrared spectrometer and its accessories. It is equipped with an integrating sphere diffuse reflectance accessory for spectral acquisition of solid samples (active pharmaceutical ingredients, contents of finished capsules), and an intelligent transmission liquid accessory (equipped with a 50mm diameter cuvette) for spectral acquisition of liquid samples (aqueous extracts, alcoholic extracts, concentrated solutions). Spectral acquisition conditions: Scanning range 10000~4000 cm⁻¹ -1 8 cm resolution -1 Each sample was scanned three times, with 32 scans per scan, and the average spectrum was taken as the final spectrum. The instrument was preheated for 30 minutes after powering on to ensure the light source was stable.

[0180] Storage module: Employs the hard disk drive of an industrial control computer, storing a pre-established near-infrared spectral quantitative calibration model library for the entire process of Shuangshitonglin capsules. The model library specifically includes: (1) Correction models for raw materials: Cortex Phellodendri correction model (quality parameters: content of palmatine hydrochloride, content of berberine hydrochloride, moisture, color) and Cortex Salvia miltiorrhiza correction model (quality parameters: content of rosmarinic acid, content of salvianolic acid B, content of cryptotanshinone, content of tanshinone I, content of tanshinone IIA, moisture, color); (2) Intermediate correction models during production: Phellodendron bark water extract-concentrate model (quality parameters: content of palmatine hydrochloride and berberine hydrochloride), Tanshinone alcohol extract-concentrate model (quality parameters: content of rosmarinic acid, salvianolic acid B, cryptotanshinone, tanshinone I, and tanshinone IIA), and mixed water extract-concentrate model of six Chinese herbal medicines (quality parameters: content of genipin, chlorogenic acid, verbascoside, and salvianolic acid B). (3) Finished product calibration model: Shuangshitonglin capsule contents model (quality parameters: content of verbascoside, tanshinone B, berberine hydrochloride, tanshinone IIA, as well as moisture and color).

[0181] For the methods of establishing each model, please refer to the specific implementation section of this manual. The model evaluation indicators Rc and Rp are both greater than 0.85, and RMSEC and RMSEP are both less than 50% of the standard deviation of the corresponding quality parameter reference values.

[0182] The calculation module uses the central processing unit of an industrial control computer to run TQ Analyst 9.0 or MATLAB R2021b analysis software. It is used to preprocess the near-infrared spectrum input from the acquisition module, call the corresponding calibration model in the storage module, and use partial least squares method to calculate the mass parameters of the sample to be tested.

[0183] Output module: It uses the display and data interface of an industrial control computer to output the quality parameters obtained by the calculation module. The output forms include real-time display on the display, printing test reports by printer, and uploading to the enterprise's production process execution system (MES) or laboratory information management system (LIMS) through the data interface.

[0184] The application process of this system in the actual production of Shuangshitonglin capsules is as follows: (1) Inspection of raw materials upon arrival at the factory After the arrival of the raw materials, Phellodendron bark or Salvia miltiorrhiza slices are sampled, pulverized through a No. 5 sieve (80 mesh), and their near-infrared spectra are acquired by the near-infrared spectroscopy acquisition module. The calculation module calls the Phellodendron bark or Salvia miltiorrhiza calibration model stored in the storage module to automatically calculate the content of indicator components, moisture, and color values ​​in this batch of raw materials. The output module displays the test results and compares them with the preset internal control standards to determine whether the raw materials are qualified, achieving rapid release. The entire process from sampling to output of results for a single sample does not exceed 5 minutes.

[0185] (2) Online monitoring of extraction / concentration during production process A flow-through cell is installed on the circulation pipeline of the *Phellodendron amurense* extraction tank, the *Salvia miltiorrhiza* alcohol extraction tank, and the six-herb mixed extraction tank. The transmission liquid accessory of the near-infrared spectroscopy acquisition module is connected to the flow-through cell. During extraction and concentration, the near-infrared spectrum of the extract or concentrate is automatically acquired every 2-6 minutes. The calculation module automatically calls the *Phellodendron amurense* aqueous extract / concentrate model, the *Salvia miltiorrhiza* alcohol extract / concentrate model, or the six-herb mixed aqueous extract / concentrate model based on the sample source to calculate the content of each indicator component in real time. The output module displays the content change curve in real time as a trend graph. When the content reaches the preset endpoint threshold, the system automatically prompts the extraction or concentration endpoint and can feed back to the equipment control system via the data interface, achieving automatic endpoint determination.

[0186] (3) Finished capsule factory inspection After capsule filling, a sample of the capsule contents is taken, and its near-infrared spectrum is acquired by the near-infrared spectroscopy acquisition module. The calculation module calls the finished product calibration model to calculate the content of four components—vermiculin, salvianolic acid B, berberine hydrochloride, and tanshinone IIA—as well as the moisture and color values. The output module generates a finished product quality inspection report for factory release determination. The full range of quality parameters for a single capsule sample can be tested within 2 minutes, significantly improving detection efficiency compared to traditional high-performance liquid chromatography (which requires 30-60 minutes for a single indicator).

[0187] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for full-process quality testing of Shuangshitonglin capsules, characterized in that, Includes the following steps: (1) Establish a quantitative calibration model for near-infrared spectroscopy of Shuangshitonglin capsules throughout the entire process. The model includes a calibration model for raw materials, a calibration model for intermediates during the production process, and a calibration model for finished products. (2) Near-infrared spectra of the raw material sample, intermediate sample during production, and contents sample of the finished capsule were collected respectively; (3) Input the near-infrared spectra collected in step (2) into the corresponding calibration models in step (1) to calculate the mass parameters of each sample to be tested.

2. The detection method according to claim 1, characterized in that, The active pharmaceutical ingredient calibration model includes the Phellodendron bark calibration model and the Salvia miltiorrhiza calibration model; The Guanhuangbai calibration model uses Guanhuangbai medicinal material as the detection object, and its quality parameters include palmatine hydrochloride content, berberine hydrochloride content, moisture and color. The Danshen calibration model uses Danshen medicinal material as the detection object, and its quality parameters include rosmarinic acid content, salvianolic acid B content, cryptotanshinone content, tanshinone I content, tanshinone IIA content, moisture and color.

3. The detection method according to claim 1, characterized in that, The intermediate correction models in the production process include the Phellodendron amurense aqueous extract model, the Phellodendron amurense concentrated extract model, the Salvia miltiorrhiza alcohol extract model, the Salvia miltiorrhiza concentrated extract model, the mixed aqueous extract model of six Chinese herbal medicines, and the mixed concentrated extract model of six Chinese herbal medicines. The detection objects of the Guanhuangbai aqueous extract model and Guanhuangbai concentrated extract model are Guanhuangbai aqueous extract and concentrated extract, and the quality parameters are the contents of palmatine hydrochloride and berberine hydrochloride. The detection objects of the tanshinone ethanol extract model and the tanshinone concentrate model are tanshinone ethanol extract and concentrate, and the quality parameters are the contents of rosmarinic acid, salvianolic acid B, cryptotanshinone, tanshinone I and tanshinone IIA. The detection objects of the six-herb mixed aqueous extract model and the six-herb mixed concentrated liquid model are the mixed aqueous extract and concentrated liquid composed of Piper longum, Patrinia scabiosaefolia, Plantago asiatica, talc, Poria cocos, and Salvia miltiorrhiza. The quality parameters are the contents of genipin, chlorogenic acid, verbascoside, and salvianolic acid B.

4. The detection method according to claim 1, characterized in that, The finished product calibration model uses the contents of Shuangshitonglin capsules as the test object, and its quality parameters include verbascoside content, salvianolic acid B content, berberine hydrochloride content, tanshinone IIA content, as well as moisture and color.

5. The detection method according to any one of claims 1-4, characterized in that, The near-infrared spectroscopy acquisition conditions are: spectral scanning range 10000~4000 cm⁻¹ -1 8cm resolution -1 Each sample was scanned 2-3 times, with each scan consisting of 32 passes.

6. The detection method according to claim 1, characterized in that, The method for establishing the calibration model in step (1) includes the following sub-steps: (a) Collect representative samples and determine the reference values ​​of the sample quality parameters using high performance liquid chromatography or conventional physicochemical analysis methods; (b) Acquire the near-infrared spectrum of the sample; (c) Preprocessing the acquired spectra, wherein the preprocessing method is selected from one or more combinations of standard normal variable transformation, multivariate scattering correction, derivative and smoothing techniques; (d) Select the spectral modeling interval and the number of principal factors, and use partial least squares method to establish a quantitative correction model between near-infrared spectra and mass parameter reference values.

7. The detection method according to claim 6, characterized in that, In the spectral preprocessing method: For the model of berberine hydrochloride in Phellodendron amurense raw material, the Constant + 1st D + SG pretreatment method was used; for the model of palmatine hydrochloride in Phellodendron amurense raw material, the Constant + 2st D + SG pretreatment method was used; for the model of color in Phellodendron amurense raw material, the Constant + 1st D + SG pretreatment method was used; for the model of moisture in Phellodendron amurense raw material, the Constant + 2st D + ND pretreatment method was used. For the models of rosmarinic acid and salvianolic acid B in tanshinone raw material, the SNV+2stD+ND pretreatment method was used; for the model of tanshinone IIA in tanshinone raw material, the MSC+1stD+ND pretreatment method was used; for the model of cryptotanshinone in tanshinone raw material, the MSC+1stD+SG pretreatment method was used; for the model of tanshinone I in tanshinone raw material, the MSC+2stD+SG pretreatment method was used; and for the models of moisture and color in tanshinone raw material, the SNV pretreatment method was used. For the models of berberine hydrochloride and palmatine hydrochloride in the aqueous extract of Phellodendron chinense, the MSC+2stD+ND pretreatment method was used; for the models of palmatine hydrochloride and berberine hydrochloride in the concentrated extract of Phellodendron chinense, the Constant+2stD+ND pretreatment method was used. For the models of each component in the ethanol extract of tanshinone, the MSC + 2stD + ND pretreatment method was used; for the model of salvianolic acid B in the tanshinone concentrate, the MSC pretreatment method was used; for the model of tanshinone IIA in the tanshinone concentrate, the MSC + 1stD + ND pretreatment method was used; for the models of rosmarinic acid, cryptotanshinone, and tanshinone I in the tanshinone concentrate, the SNV pretreatment method was used. For the model of salvianolic acid B in the mixed aqueous extract of six Chinese herbs, the MSC + 1stD + ND pretreatment method was used; for the models of genipin and verbascoside in the mixed aqueous extract of six Chinese herbs, the MSC pretreatment method was used; for the model of chlorogenic acid in the mixed aqueous extract of six Chinese herbs, the MSC + 2stD + ND pretreatment method was used; for the models of verbascoside and salvianolic acid B in the mixed concentrated extract of six Chinese herbs, the Constant + 1stD pretreatment method was used. For the model of genipinic acid in the concentrated solution of six Chinese herbal medicines, the MSC+1stD pretreatment method was used; For the model of chlorogenic acid in the concentrated solution of six Chinese herbal medicines, the MSC+2stD+SG pretreatment method was used; For the models of verbascoside and salvianolic acid B in the finished product of Shuangshitonglin capsules, the SNV+1stD pretreatment method was used; for the model of berberine hydrochloride in the finished product, the MSC+2stD+SG pretreatment method was used; for the model of tanshinone IIA in the finished product, the MSC+1stD pretreatment method was used; for the model of moisture in the finished product, the Constant+1stD+ND pretreatment method was used; and for the model of color in the finished product, the Constant+1stD pretreatment method was used.

8. A system for implementing the detection method according to any one of claims 1-7, characterized in that, include: The near-infrared spectroscopy acquisition module is used to acquire near-infrared spectra of active pharmaceutical ingredients, intermediates, and finished product samples; the storage module stores a pre-established calibration model for the active pharmaceutical ingredient, a calibration model for the intermediates during production, and a calibration model for the finished product, according to any one of claims 1-7; the calculation module is used to input the acquired spectra into the corresponding calibration models and calculate the quality parameters; and the output module is used to output the calculated quality parameters.

Images