On-line component detection method and system based on chromatography elution process of salvia miltiorrhiza extracting solution

Abstract

The invention discloses an online component detection method and system based on a salvia miltiorrhiza extracting solution chromatography elution process, and relates to the technical field of traditional Chinese medicine quality control. The method comprises the following steps: carrying out an elution experiment on a salvia miltiorrhiza extracting solution based on a column chromatography technology to obtain a sample set; acquiring near infrared spectrum and ultraviolet visible spectrum data in the sample; matching a spectrum fusion model according to the type of the to-be-detected component, and constructing different spectrum fusion models based on different fusion strategies for the near infrared spectrum and the ultraviolet visible spectrum; and processing data of the near infrared spectrum and the ultraviolet visible spectrum by using different spectrum fusion models to obtain the content of the to-be-detected component. On the basis of NIRS and UV-Vis technologies, online detection and analysis are carried out in the elution process of the salvia miltiorrhiza extracting solution by adopting a spectrum fusion and water spectrum omics method, and dynamic tracking and intelligent terminal point judgment of the salvia miltiorrhiza water-soluble active ingredients are realized.

Description

Technical Field

[0001] This invention relates to the field of traditional Chinese medicine quality control technology, and in particular to an online detection method and system for components based on the chromatographic elution process of Danshen extract. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] Salvia miltiorrhiza B.G. is the dried root and rhizome of the plant Salvia miltiorrhiza, belonging to the Lamiaceae family. It is bitter and slightly warm in nature, and enters the heart and liver meridians. Studies have shown that Salvia miltiorrhiza and its extracts exhibit significant benefits in the prevention and treatment of heart and kidney diseases such as atherosclerosis and chronic kidney disease.

[0004] Modern biological, chemical, and pharmacological studies have shown that the traditional Chinese medicine (TCM) summary of the efficacy of Danshen (Salvia miltiorrhiza) is based on its chemical components. As one of the earliest studied TCM herbs, Danshen's research model has been helpful for the research and development of other medicines. However, due to the complexity of TCM components, the dose-effect relationship of their effects is still in the preliminary research stage. Further in-depth research is needed on the active components of Danshen, the interrelationships among various chemical components, and the mechanisms of action of each chemical component. Therefore, the quality control and evaluation of TCM materials is one of the core issues restricting the modernization of TCM, and it is also a highly concerned and challenging area in modern TCM research. With the advancement of modern science and technology, the quality evaluation system for TCM has developed rapidly, and the integration of multiple fields and technologies has provided new possibilities for the high-quality development of TCM.

[0005] Modern quality evaluation of traditional Chinese medicine employs techniques such as gene identification, chemical composition analysis, biological effect evaluation, and control of exogenous contaminants, establishing a relatively comprehensive quality control system. Particularly in the field of pharmaceutical engineering, the combination of chemical composition analysis and process analysis techniques allows for real-time monitoring of component changes during production, providing a scientific basis for the quality control of traditional Chinese medicine and its products.

[0006] As a traditional Chinese medicine, the quality evaluation methods for Danshen (Salvia miltiorrhiza) and its related products are also applicable to chemical component analysis. In studies on the correlation between the medicinal components and properties of drugs in the Lamiaceae family, the most closely related components are salvianolic acid A, salvianolic acid B, shikonin, and tanshinone I, tanshinone IIA, among others, and their cold nature and bitter taste. In the production process of Danshen-related products, the content of salvianolic acid A, salvianolic acid B, tanshinone IIA, cryptotanshinone, and tanshinone I can serve as key indicators for quality evaluation. Among these, salvianolic acid A and salvianolic acid B are water-soluble components in Danshen, while tanshinone IIA, cryptotanshinone, and tanshinone I are fat-soluble components. Comprehensive analysis of these components allows for a complete assessment of the quality characteristics of Danshen products, providing a scientific basis for quality control in its production process.

[0007] Salvianolic acid A is one of the water-soluble components of Salvia miltiorrhiza, and its chemical structural formula is as follows: Figure 1 As shown, it can be considered as a condensation of two molecules of tanshinone and one molecule of caffeic acid. Tanshinone A is one of the most active water-soluble components in tanshinone, exhibiting strong anti-lipid peroxidation and free radical scavenging effects. Due to its broad pharmacological activity, tanshinone A has broad application prospects in health foods and pharmaceuticals. Since its discovery, tanshinone A has been one of the hot topics in the research of active components of tanshinone. However, the content of tanshinone A in tanshinone is relatively low (approximately 0.1‰), far lower than other water-soluble phenolic acid components, such as tanshinone B (4%), tanshinone (8‰), tanshinone C (1.5‰), tanshinone D (0.06‰), shikonin (30‰), and rosmarinic acid (20‰). Due to its difficulty in obtaining large quantities and high price, the further development and application of tanshinone A in cosmetics, food, and pharmaceuticals are limited. Therefore, how to efficiently obtain tanshinone A has become a hot research topic in the fields of biology, pharmacy, and chemical synthesis in recent years.

[0008] Salvianolic acid B is an important water-soluble component of Salvia miltiorrhiza, one of the most abundant and bioactive components in the plant, a key indicator for quality evaluation, and one of the most extensively studied compounds among its active ingredients. Its chemical structural formula is as follows: Figure 2 As shown, it can be considered to be formed by the condensation of three molecules of tanshinone and one molecule of caffeic acid.

[0009] Water-soluble active ingredients in Danshen (such as salvianolic acid A and salvianolic acid B) have significant value in the treatment of cardiovascular diseases, but their industrial production process urgently requires efficient online detection technology to overcome the limitations of traditional high-performance liquid chromatography (HPLC). Traditional Danshen elution processes mainly rely on column chromatography to separate active ingredients such as salvianolic acid A and salvianolic acid B. Detection in this process primarily employs offline analysis techniques, mainly HPLC, which are time-consuming and require highly skilled operators, making real-time feedback difficult. With increasingly stringent drug regulations, the widespread adoption of intelligent manufacturing concepts, and the rapid development of process analysis technology (PAT), analytical techniques such as near-infrared spectroscopy (NIRS) and ultraviolet-visible spectroscopy (UV-Vis) are gradually becoming important tools for drug production process monitoring and product quality control due to their advantages of speed, non-destructive nature, and environmental friendliness. However, the application of a single rapid detection technology often suffers from problems such as the inability to provide comprehensive information on molecular vibrational modes, limitations in detection methods, and limited analytical capabilities for complex systems.

[0010] Therefore, how to achieve real-time, rapid, and efficient detection of water-soluble active ingredients in Danshen has become a technical problem that urgently needs to be solved by existing technologies. Summary of the Invention

[0011] To address the shortcomings of existing technologies, the present invention aims to provide an online detection method and system for components based on the chromatographic elution process of Salvia miltiorrhiza extract. Based on NIRS and UV-Vis technologies, the method employs spectral fusion and aqueous spectroscopic omics to perform online detection and analysis during the elution process of Salvia miltiorrhiza extract, thereby achieving dynamic tracking and intelligent endpoint determination of water-soluble active components of Salvia miltiorrhiza.

[0012] To achieve the above objectives, the present invention is implemented through the following technical solution: The first aspect of this invention provides an online method for component detection based on the chromatographic elution process of Salvia miltiorrhiza extract, comprising the following steps: Elution experiments were conducted on the extract of Salvia miltiorrhiza using column chromatography to obtain a sample set. Acquire near-infrared and ultraviolet-visible spectral data of the sample; The spectral fusion model is matched according to the type of component to be detected, and different spectral fusion models are constructed based on different fusion strategies for near-infrared and ultraviolet-visible spectra. Different spectral fusion models were used to process near-infrared and ultraviolet-visible spectral data to obtain the content of the component to be detected.

[0013] A second aspect of the present invention provides an online component detection system based on the chromatographic elution process of Salvia miltiorrhiza extract, comprising: The experimental module is configured to perform elution experiments on the extract of Salvia miltiorrhiza based on column chromatography to obtain a sample set; The data acquisition module is configured to acquire near-infrared and ultraviolet-visible spectral data of the sample; The model matching module is configured to match spectral fusion models according to the type of the component to be detected, wherein different spectral fusion models are constructed based on different fusion strategies for near-infrared and ultraviolet-visible spectra. The online detection module is configured to process near-infrared and ultraviolet-visible spectral data using different spectral fusion models to obtain the content of the component to be detected.

[0014] A third aspect of the present invention provides a computer-readable storage medium storing a computer program adapted to be loaded by a processor and to execute steps in the online component detection method based on the chromatographic elution process of tanshinone extract as described in the first aspect of the present invention.

[0015] A fourth aspect of the present invention provides a computer device comprising: A processor, adapted to execute computer programs; A computer-readable storage medium storing a computer program, which, when executed by the processor, implements the online component detection method based on the chromatographic elution process of Danshen extract as described in the first aspect of the present invention.

[0016] The above one or more technical solutions have the following beneficial effects: This invention discloses an online detection method and system for components based on the chromatographic elution process of Tanshinone extract. Based on a laboratory simulation of the elution process for salvianolic acid production, an online detection system combining ultraviolet-visible (UV-Vis) and near-infrared (NIR) spectroscopy was successfully constructed, enabling real-time dynamic tracking of the concentrations of salvianolic acid A and B during the elution process of Tanshinone extract. This invention utilizes a combination of UV and NIR detection methods for online detection of the Tanshinone extract elution process, aiming to achieve complementary advantages and thus a superior online detection effect for the entire elution process. Accurate and comprehensive detection can be obtained at different stages of the elution process, whether at the initial low concentration or the potentially high concentration later. The research results provide a theoretical basis and technical path for multispectral fusion modeling, laying a methodological foundation for the rapid analysis of complex natural products. The UV-Vis-NIR multispectral coupling not only makes up for the shortcomings of NIR in the detection of trace components, but also enhances the analytical capability of complex systems (such as the coexistence and interference of polysaccharides and tannins in Danshen extract) through multidimensional information integration, providing efficient and reliable technical support for real-time control and quality optimization of the elution process in the production of traditional Chinese medicine.

[0017] This invention aims to construct an online detection system to achieve real-time detection and endpoint determination of salvianolic acid A and B in tanshinone extract during the elution process. A multispectral fusion model is constructed by combining NIRS and UV-Vis to monitor the content of salvianolic acid A and B during elution. Simultaneously, a qualitative analysis model is established using the principles of NIRS-based aqueous spectroscopic omics to accurately determine the elution endpoint of salvianolic acid A. The implementation of this invention will help improve the intelligent manufacturing level of tanshinone-related products, ensure the stability and consistency of product quality, thereby promoting the modernization of traditional Chinese medicine preparation production and providing technical support for the digital transformation of the traditional Chinese medicine industry.

[0018] Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a schematic diagram of the chemical structure of salvianolic acid A in this invention; Figure 2This is a schematic diagram of the chemical structure of salvianolic acid B in this invention; Figure 3 This is a flowchart of the online component detection method based on the chromatographic elution process of tanshinone extract in Embodiment 1 of the present invention; Figure 4 This is a schematic diagram of the fusion strategy in Embodiment 1 of the present invention. Detailed Implementation

[0021] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0022] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof. The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0023] Example 1: Embodiment 1 of the present invention provides an online component detection method based on the chromatographic elution process of tanshinone extract, such as... Figure 3 As shown, taking the detection of salvianolic acid A and salvianolic acid B as an example, the following steps are included: Step 1: Elution experiment of Danshen extract was carried out based on column chromatography to obtain sample set.

[0024] Column chromatography, also known as column chromatography, is a form of chromatography. The stationary phase consists of a porous solid within a cylindrical transparent tube. The mobile phase carrying the analyte mixture flows through the stationary phase, gradually separating the mixture. Major types include adsorption column chromatography, gel column chromatography, and ion exchange column chromatography.

[0025] Step 1.1: Prepare for the elution experiment.

[0026] The preparation work for the elution experiment mainly involves the pretreatment of the macroporous resin and column packing. D101 macroporous resin was soaked in 10 times its volume of 95% ethanol for 24 hours, then wet-packed into the column. It was washed with 95% ethanol until the eluent, when mixed with an appropriate amount of water, showed no white turbidity. This eluent was then added to a 16 mm 20 cm chromatography column and allowed to stand. Finally, distilled water was added and washed until no alcohol odor was detected. The eluent was prepared by mixing 95% ethanol with 20% and 40% ethanol solutions.

[0027] Step 1.2: Build a sample and data fusion spectral acquisition system to conduct elution experiments and obtain samples to form a sample set.

[0028] The experimental procedure described in this embodiment draws upon the elution process in industrial production, and a sample and data fusion spectral acquisition system was constructed. The experimental procedure involves preparing ethanol solutions of different concentrations and passing them through macroporous resin containing concentrated tanshinone extract to elute and collect salvianolic acid A and salvianolic acid B. Throughout the experiment, the resin-packed chromatography column was fixed on a suitable support and connected to a peristaltic pump. First, the peristaltic pump was used to control the flow rate, allowing five column volumes of water to slowly pass through the chromatography column. The purpose was to clean the column and macroporous resin, removing any potential impurities, and simultaneously allowing the resin to fully swell, creating favorable conditions for subsequent sample separation. During the water flow, the flow of liquid within the chromatography column was closely observed to ensure there was no blockage or leakage. After water washing, the peristaltic pump was used to pass five column volumes of 20% ethanol through the chromatography column at a stable flow rate. During this process, some impurities and components with weak binding to the macroporous resin were eluted along with the 20% ethanol. After elution with 20% ethanol was complete, elution was switched to 40% ethanol.

[0029] Based on previous experiments and related research, the elution range between 1 volume of 40% ethanol and 5 volumes of 40% ethanol was determined to be the elution interval for salvianolic acid A and salvianolic acid B. During the elution process, samples were collected at intervals of 0.2 column volumes.

[0030] The specific procedures for the experiment are as follows: The peristaltic pump was started to allow 40% ethanol to flow into the chromatography column. When the eluent volume reached one column volume, sample collection began. One sample was collected in a clean, numbered container after every 0.2 column volumes of eluent, and refrigerated. During collection, the collection order and corresponding eluent volume of each sample were accurately recorded to ensure accuracy and traceability. Collected samples should be properly stored. Based on the properties of salvianolic acid A and B, they should be refrigerated during the experiment. After each batch of experiments, they should be stored at -20°C to prevent degradation or alteration of the components, pending subsequent analysis. A total of six batches of samples were obtained using this method, forming a sample set.

[0031] Step 2: Obtain near-infrared and ultraviolet-visible spectral data of the sample.

[0032] In addition to elution experiments, the sample and data fusion spectral acquisition system can also perform spectral acquisition. Specifically, the light source is located on the right side, and the light emitted from the light source enters the flow cell via an optical fiber connection. The eluted sample is also driven through the flow cell by a pump. The other side of the flow cell uses a Y-shaped optical fiber, providing two sensors that acquire the spectrum of the eluted sample via the Y-shaped fiber. These are a portable OTO NIRS acquisition system and a miniature UV-vis sensor, respectively acquiring near-infrared and ultraviolet-visible spectra to obtain relevant sample information. The flow cell contains an optical fiber probe and corresponding optical path structure. The sample is driven through the flow cell by the pump, and the light source and optical fiber probe are used to emit and collect light signals. Then, different detection devices analyze the spectral signals, which can be used for real-time detection of sample composition, concentration, and other characteristics.

[0033] During elution, samples were collected at intervals of 0.2 column volumes. Spectral acquisition was set to occur every 30 seconds. The specific procedure for all spectral experiments within 0.2 column volumes was as follows: the peristaltic pump was started to allow 40% ethanol to flow into the chromatography column. When the eluent volume reached one column volume, sample collection began. The average of the spectra obtained after each 0.2 column volume eluent was used as the original data for model establishment. A total of 73 industrial-level spectral data samples were obtained using this method.

[0034] Step 3: Match the spectral fusion model according to the type of component to be detected.

[0035] Among them, different spectral fusion models are constructed based on different fusion strategies for near-infrared and ultraviolet-visible spectra.

[0036] In one specific implementation, different fusion strategies are constructed based on the spectral characteristics of salvianolic acid A and salvianolic acid B, including primary fusion strategies, intermediate fusion strategies, and advanced fusion strategies, such as... Figure 4 As shown.

[0037] The specific steps of the initial fusion strategy include: Near-infrared and ultraviolet-visible spectral data are normalized by mean and then spliced ​​in a certain order to form a new dataset containing information from both spectral types. The PLS method is then used to build a model based on this dataset, aiming to retain the most comprehensive original data information.

[0038] The specific steps of the intermediate fusion strategy include: Principal component analysis (PCA) was used to extract features from near-infrared (NIIR) and ultraviolet-visible (UV-Vis) spectral data. The extracted features yielded the NIIR feature matrix (TUV-Vis) and the UV-Vis feature matrix (TNIR). Subsequently, the feature data were selectively fused based on the proportion of each principal component. Specifically, weights were dynamically assigned based on the variance contribution rate of the principal components to achieve feature-level fusion. The optimal fusion model was then constructed and selected. Specifically, different combinations of principal component numbers (e.g., UV-LVs = 1~5, NIR-LVs = 1~5) were iterated, and the combination that minimized RMSEP and achieved RPD > 2.5 was selected as the optimal fusion model. This approach enables efficient integration of data features, reduces data dimensionality, and improves model training efficiency.

[0039] It should be noted that the features extracted by the PCA method in this embodiment do not directly correspond to chemical components, but are latent variables (LVs) mined from the original spectra, namely the co-fluctuation signals related to concentration changes in UV-Vis and NIR spectra.

[0040] The specific steps of the advanced integration strategy include: Advanced fusion builds upon primary fusion by constructing separate classification or prediction models using UV-Vis and NIR data. Then, based on the prediction results of different models, a decision-level fusion is performed using a stacking method to form the final fusion model.

[0041] The specific steps of this embodiment to perform decision-level fusion using a stacking method to form the final fusion model are as follows: First, build and train the base models, establishing two base models: Model_UV-Vis and Model_NIR.

[0042] 1. Dataset partitioning: The sample set is divided into a training set (70%) and a validation set (30%).

[0043] 2. Establish Model_UV-Vis and Model_NIR, and train the base model using the training set. The Model_UV-Vis model is obtained by training a PLS regression model using only UV-Vis spectral data. The Model_NIR model is obtained by training a PLS regression model using only NIR spectral data.

[0044] 3. Validate the training results on the validation sets Model_UV-Vis and Model_NIR, and output the predicted values ​​of the Model_UV-Vis and Model_NIR models.

[0045] Next, a meta-dataset is constructed, combining the predictions from the Model_UV-VIS and Model_NIR models into new features. The formula for constructing the meta-dataset is: , .

[0046] in, For the spectral meta-dataset, The data are ultraviolet-visible spectral data. Near-infrared spectral data, The true value refers to the actual concentration of the target component (such as salvianolic acid B) in the sample. n is the number of samples.

[0047] This formula uses the prediction results of ultraviolet-visible spectral data and near-infrared spectral data as new features to train the final meta-model.

[0048] Finally, the new features were used to train the meta-model, and the PLS model, which is simple and resistant to overfitting, was selected as the meta-model. .

[0049] in, This is a meta-model trained using the Partial Least Squares (PLS) regression algorithm. Meta-model The result of prediction on the meta-dataset is the final concentration prediction value after fusion by stacking method.

[0050] Model fusion can fully leverage the strengths of different spectral data in different models, enhancing the model's generalization ability and prediction accuracy. Examples include RMSEC, RMSEP, and R... 2 cal and R 2 p is used as an evaluation index, with smaller RMSE indicating better R. 2 Following the principle that the closer the result is to 1, the optimal model fusion method was selected for salvianolic acid A and B respectively. Based on experimental performance analysis, the effectiveness of fusion strategies based on salvianolic acid A content was ranked as follows: intermediate fusion strategy > primary fusion strategy > advanced fusion strategy; the effectiveness of fusion strategies based on salvianolic acid B content was ranked as follows: advanced fusion strategy > intermediate fusion strategy > primary fusion strategy. Therefore, in this embodiment, the spectral fusion model constructed using the intermediate fusion strategy was used to detect the content of salvianolic acid A, and the spectral fusion model constructed using the advanced fusion strategy was used to detect the content of salvianolic acid B.

[0051] Step 4: Process the near-infrared and ultraviolet-visible spectral data using different spectral fusion models to obtain the content of the component to be detected.

[0052] Step 4.1: Preprocess the near-infrared and ultraviolet-visible spectral data.

[0053] The portable device used outputs spectral data as light intensity data. Therefore, before building the model, the light intensity data of the sample spectrum must be converted into absorbance. This requires subtracting the effects of background and dark current before calculation. The specific formula for subtracting the effect of dark current from the background spectrum and sample spectrum is as follows: (1), (2).

[0054] in, The background light intensity after deducting dark current. Background light intensity, Dark current light intensity, The sample spectrum after dark current subtraction. The light intensity is the sample light intensity.

[0055] Step 4.2: Convert the preprocessed near-infrared and ultraviolet-visible spectral data into absorbance.

[0056] Transmittance T is the ratio of sample-corrected light intensity to background-corrected light intensity: (3).

[0057] Absorbance A is calculated using the negative logarithm of transmittance (base 10): (4). Step 4.3: Use the matched spectral fusion model to fuse the absorbance data.

[0058] The modeling process employed MATLAB 2020a software to process the raw near-infrared and ultraviolet-visible spectra collected during the elution of *Salvia miltiorrhiza* extract, establishing quantitative analytical models for salvianolic acid A and B. Specifically, HPLC was used to obtain the content information of salvianolic acid A and B in the eluent, which was then used as the primary data from the near-infrared spectra. In this embodiment, a total of 73 samples from 6 batches of industrial-level spectral data were collected for modeling.

[0059] Step 4.3.1: Remove outliers from the absorbance data.

[0060] Before performing PLS modeling, PCA is used to remove outliers from the experimentally collected NIR spectral data, which can eliminate the interference of abnormal spectra or samples on the model.

[0061] Specifically, with a significance level of α=0.05, by observing the principal component scores and further combining them with the contributions of the variables, we analyze the degree of contribution of each variable in the principal components, and identify the sample data corresponding to variables that have a significant impact on the principal components and exhibit abnormal fluctuations. We then calculate Hotelling's T for each sample. 2 The statistical measure is used to classify samples that exceed a critical value as outliers. In this example, two samples that fell within the 95% confidence interval were removed, leaving 71 samples for subsequent analysis.

[0062] Step 4.3.2: Perform fusion analysis on the absorbance data after removing outliers.

[0063] The initial fusion modeling involved performing MC processing on the UV-Vis and NIR spectra of salvianolic acid A and salvianolic acid B respectively. Then, the data from the two different sources were linked according to the wavelength order from UV-Vis to near-infrared. After data fusion, a new PLS model was established. The model results are shown in Table 1. After directly fusing the UV-Vis and NIR spectra, the RL of salvianolic acid A... 2 The p-value increased to 0.9387, and the RPD also significantly improved to 2.8034, indicating that the combined use of UV-Vis and NIR spectroscopy can effectively enhance the model's characterization ability of salvianolic acid A, and the performance of the salvianolic acid A model under primary fusion is significantly improved. The R-value of salvianolic acid B... 2 The p-value was 0.9132, and the RPD was 2.2449, slightly lower than the standalone UV-Vis model (Rp). 2 (p=0.9334). Compared to the standalone UV-Vis model, primary fusion performed better on salvianolic acid A, but slightly worse on salvianolic acid B. This may indicate that primary fusion is more effective for salvianolic acid A, effectively integrating the complementary information of UV-Vis and NIR, resulting in better prediction performance than a single spectrum, thus validating the value of multispectral fusion. For the primary fusion model of salvianolic acid B, although RPD=2.2449 is still practical, R... 2 Since p has decreased, we will further explore the potential value of NIR data and improve the analytical capabilities for complex samples by changing and integrating strategies (such as feature selection). To verify the model's low-concentration prediction effect, the results of the initial fusion models for salvianolic acid A and B are shown in Figure 3-11. It can be seen that the introduction of UV-Vis effectively avoids the dense occurrence of outliers in the predicted samples of low concentrations (salvianolic acid A concentration less than 0.15 mg / mL and salvianolic acid B concentration less than 0.2 mg / mL), i.e., the phenomenon of large errors.

[0064] Table 1. Modeling results of salvianolic acid A and B after initial model fusion

[0065] Intermediate fusion modeling is a data selection method that combines feature-level data fusion extracted from principal components with principal component optimization. The number of features selected in this study is based on the Levels of Features (LVs) derived from the primary modeling described in 3.3.3.2.

[0066] Table 2 shows the intermediate fusion results of salvianolic acid A, demonstrating the model performance under different combinations of principal component feature extraction numbers (UV-LVs and NIR-LVs). Specifically, when UV-LVs=3 and NIR-LVs=2, the RMSEP=0.0086 mg / mL, R 2 With p=0.9712 and RPD=3.9115, it performed best among all combinations, indicating that the feature fusion of salvianolic acid A data can extract complementary information between UV-Vis and NIR, significantly improving prediction accuracy. As the number of NIR-LVs decreases (e.g., from 5 to 2), RPD... 2 The p and RPD values ​​fluctuate but have an overall upward trend, indicating that appropriately reducing the principal components of NIR may help avoid overfitting. NIR provides information on molecular skeleton or solvent effects through CH bond vibrations and provides effective information supplementation in low dimension, but noise needs to be eliminated through principal component screening (e.g., LVs-NIR=2).

[0067] Table 2 Modeling results of salvianolic acid A after intermediate fusion in the model

[0068] Table 3 shows the intermediate-level fusion results of salvianolic acid B, also demonstrating the performance indicators under different combinations of principal component extraction numbers. Through principal component characteristic-level joint screening, the combination with UV-LVs=9, NIR-LVs=6, RMSEP=0.0209 mg / mL, and R... 2 p=0.9341, RPD=2.4744, the highest values ​​among all intermediate-level fusion modeling models. Compared with the modeling results of salvianolic acid B in the primary fusion, the intermediate-level fusion optimized feature expression through principal component screening, but the improvement was limited. Adjusting the number of principal components failed to significantly improve performance, reflecting insufficient data quality or chemical information complementarity, which may be related to the weak NIR signal contribution of salvianolic acid B. Furthermore, the best model performance of the intermediate-level fusion of salvianolic acid B (RMSEP=0.0209 mg / mL, RPD=2.4744) was the highest. 2 (p=0.9341) did not outperform the standalone UV-Vis spectral model of salvianolic acid B. This demonstrates that intermediate-level fusion has reached its performance limit, and further breakthroughs are unlikely.

[0069] Table 3 Modeling results of salvianolic acid B after intermediate fusion in the model

[0070] The intermediate-level fusion results are summarized as follows: the model for salvianolic acid A was significantly optimized. Complementary information was effectively extracted through principal component screening (UV-LVs=3, NIR-LVs=2), with a prediction performance RMSEP of 0.0086 mg / mL and R... 2 p=0.9712, RPD=3.9115, validating the value of intermediate-level fusion. The intermediate-level fusion model for Tanshinone B is limited by NIR data quality and UV-Vis performance saturation. Without other fusion methods, preprocessing (such as noise suppression and feature enhancement) or band optimization is necessary to unlock model performance. Optimizing the number of principal components is key to balancing model complexity and performance, but it requires dynamic adjustment based on the characteristics of different compounds. The strong specificity of the UV-Vis spectrum (conjugated structure response) remains fundamental to improving model performance.

[0071] Advanced fusion modeling involves integrating the prediction results of non-homologous models for the same target substance into a single model. In this embodiment, it is represented by using the predicted y-values ​​as the dataset for modeling. Table 4 shows the advanced fusion model results for salvianolic acid A and B.

[0072] The RMSEP of the high-order fusion model results for salvianolic acid A was 0.0159 mg / mL. 2 p=0.9393 and the best result of intermediate fusion modeling, RMSEP=0.0086 mg / mL, R 2 The p=0.9712 is relatively low, and the RPD=2.8061 differs significantly from RPD=3.9115, indicating that although the decision-making layer model fusion based on the concentration of salvianolic acid A integrates information from multiple models, it does not leverage the advantages of decision fusion.

[0073] The RMSEP of the high-order fusion model results for salvianolic acid B was 0.0187 mg / mL. 2 p=0.9381, which is higher than the RMSEP of 0.0228 mg / mL in the primary fusion modeling results. 2 The optimal result for intermediate-level fusion modeling was p=0.9132 and RMSEP=0.0221mg / mL. 2 p=0.9341 indicates that the advanced fusion based on the concentration of salvianolic acid B has a certain robustness in complex samples, and RPD=2.7362 is the best among all types of salvianolic acid B models, indicating that the model's practicality is enhanced.

[0074] Advanced fusion results indicate that the prediction results for salvianolic acid B are modeled using R... 2 The p=0.9381 value is close to the optimal value for intermediate fusion, indicating that multi-model synergy can improve the analytical capabilities of complex samples. Furthermore, the advanced fusion of Tanshinone A improves RMSEC and R by integrating multiple models. 2The model strikes a balance between accuracy and robustness, but further optimization of the algorithm and data preprocessing workflow is still needed to fully realize its potential in chemical analysis.

[0075] Table 4. Modeling results of salvianolic acid A and B after advanced fusion

[0076] This embodiment focuses on the research of a model fusion method for the concentrations of salvianolic acid A and B during the elution process of *Salvia miltiorrhiza*. By modeling NIR and UV-Vis spectra separately and by fusion modeling at different levels, the performance indicators of various indicators are compared and analyzed to explore the impact of different modeling methods on the prediction effects of salvianolic acid A and B. The effectiveness ranking of the fusion strategy based on salvianolic acid A content is: intermediate fusion > primary fusion > advanced fusion; the effectiveness ranking of the fusion strategy based on salvianolic acid B content is: advanced fusion > intermediate fusion > primary fusion.

[0077] The phenomenon of UV-Vis spectroscopy dominating model performance (such as the enhancement of salvianolic acid A in primary fusion) may be related to its greater sensitivity to electronic transition responses of phenolic acid compounds, while the vibrational spectral contribution of NIR needs to be optimized through appropriate chemometric methods. Therefore, directions for improvement include: introducing band selection or assigning UV-Vis / NIR difference weights to optimize the results of each level of fusion model for NIR data; exploring deep learning frameworks (such as attention mechanisms) or Bayesian optimization to improve the ability of advanced fusion to capture complex data patterns; and combining loading analysis with molecular structure studies to clarify the quantitative correlation between UV-Vis / NIR characteristics and phenolic acid functional groups, supporting the interpretability of the model.

[0078] This embodiment successfully constructed an online detection system based on ultraviolet-visible (UV-Vis) and near-infrared (NIR) spectroscopy, simulating the elution process of salvianolic acid production in a laboratory setting. This system enabled real-time dynamic tracking of the concentrations of salvianolic acid A and B during the elution of Salvia miltiorrhiza extract. The system integrates a flow cell, a Y-type fiber optic probe, and dual detectors (OTO NIRS analysis system and miniature UV-Vis sensor) to achieve simultaneous acquisition and fusion analysis of spectral data. Using 40% ethanol as the mobile phase, samples were collected every 0.2 column volume, and spectral data were acquired simultaneously (30 seconds / time). Primary concentration data for salvianolic acid A and B were established using HPLC, simulating 6 batches of 73 eluted samples with different concentrations.

[0079] The system optimizes model performance through primary, intermediate, and advanced fusion strategies. Primary fusion directly stitches together UV and NIR spectral data, significantly improving the prediction accuracy of salvianolic acid A (R0.05). 2 p=0.9387, RMSEP=0.0119 mg / mL); R of salvianolic acid B 2The p-value was 0.9132, RMSEP was 0.0228 mg / mL, and RPD was 2.2449, slightly lower than the standalone UV-Vis model (0.9334), indicating a performance decline. Intermediate fusion was screened using principal component features (UV-LVs=3, NIR-LVs=2), and the R-value of salvianolic acid A was... 2 p was further improved to 0.9712, RMSEP decreased to 0.0086 mg / mL, and RPD reached 3.9115, significantly optimizing the model for salvianolic acid A; using principal component feature screening (UV-LVs=9, NIR-LVs=6), the RMSEP of salvianolic acid B was further improved. 2 p increased to 0.9341, RMSEP decreased to 0.0209 mg / ml, and RPD reached 2.4744. Performance optimization of salvianolic acid B was limited, requiring further optimization using other strategies. The results of intermediate fusion validated the complementarity of UV-Vis and NIR spectra. Advanced fusion, through decision-level integration, increased the RPD of salvianolic acid B to 2.7362, overcoming the limitations of single-spectrum analysis; however, the prediction results for salvianolic acid A were poor, R… 2 p=0.9393, RMSEP=0.0159 mg / mL, RPD=2.9746, which is lower than the results of intermediate fusion with salvianolic acid A, indicating that the model with salvianolic acid A cannot give full play to the advantages of advanced fusion.

[0080] Compared to the pure NIR model (where the R2p of salvianolic acid B is only 0.3243), the above fusion model significantly reduced the dispersion of calibration points in the low concentration region (salvianolic acid A < 0.15 mg / mL, salvianolic acid B < 0.2 mg / mL), avoiding the prediction bias caused by the low signal-to-noise ratio and insufficient sensitivity of the single NIR technique. Furthermore, the system employs standardized preprocessing (dark current correction, mean centering) and outlier removal (PCA combined with Hotelling's T... 2 Statistics ensured data quality and model robustness.

[0081] The research results of this embodiment provide a theoretical basis and technical path for multispectral fusion modeling, laying a methodological foundation for the rapid analysis of complex natural products. The UV-Vis-NIR multispectral coupling not only compensates for the shortcomings of NIR in the detection of trace components, but also enhances the analytical capabilities of complex systems (such as the interference of polysaccharides and tannins in Danshen extract) through multidimensional information integration, providing efficient and reliable technical support for real-time control and quality optimization of the elution process in traditional Chinese medicine industrial production.

[0082] Example 2: Embodiment 2 of the present invention provides an online component detection system based on the chromatographic elution process of tanshinone extract, comprising: The experimental module is configured to perform elution experiments on the extract of Salvia miltiorrhiza based on column chromatography to obtain a sample set; The data acquisition module is configured to acquire near-infrared and ultraviolet-visible spectral data of the sample; The model matching module is configured to match spectral fusion models according to the type of the component to be detected, wherein different spectral fusion models are constructed based on different fusion strategies for near-infrared and ultraviolet-visible spectra. The online detection module is configured to process near-infrared and ultraviolet-visible spectral data using different spectral fusion models to obtain the content of the component to be detected.

[0083] Example 3: Embodiment 3 of the present invention provides a computer-readable storage medium storing a computer program adapted for loading by a processor and executing the steps in the online component detection method based on the chromatographic elution process of tanshinone extract as described in Embodiment 1 of the present invention.

[0084] Example 4: Embodiment 4 of the present invention provides a computer device, the device comprising: A processor, adapted to execute computer programs; A computer-readable storage medium storing a computer program, which, when executed by the processor, implements the steps in the online component detection method based on the chromatographic elution process of tanshinone extract as described in Embodiment 1 of the present invention.

[0085] The steps and methods involved in Examples 2, 3 and 4 above correspond to those in Example 1. For specific implementation details, please refer to the relevant description section of Example 1.

[0086] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed in this application can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application. In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. A computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the flow or function according to the embodiments of this application is generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in or transmitted through a computer-readable storage medium. The computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that a computer can access or a data processing device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., solid-state disk (SSD)). The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. An online method for component detection based on the chromatographic elution process of Salvia miltiorrhiza extract, characterized in that, Includes the following steps: Elution experiments were conducted on the extract of Salvia miltiorrhiza using column chromatography to obtain a sample set. Acquire near-infrared and ultraviolet-visible spectral data of the sample; The spectral fusion model is matched according to the type of component to be detected, and different spectral fusion models are constructed based on different fusion strategies for near-infrared and ultraviolet-visible spectra. Different spectral fusion models were used to process near-infrared and ultraviolet-visible spectral data to obtain the content of the component to be detected.

2. The method for online component detection based on the chromatographic elution process of Salvia miltiorrhiza extract as described in claim 1, characterized in that, The specific steps for eluting the tanshinone extract using column chromatography are as follows: Preparations for the elution experiment include pretreatment of the macroporous resin and column packing. A sample and data fusion spectral acquisition system was built to conduct elution experiments and obtain samples to form a sample set.

3. The online component detection method based on the chromatographic elution process of Tanshinone extract as described in claim 1, characterized in that, The sample and data fusion spectral acquisition system includes a light source, a flow cell, and sensors. The light emitted by the light source enters the flow cell via an optical fiber connection. The eluted sample is also driven through the flow cell by a pump. On the other side of the flow cell, two sensors acquire the spectrum of the eluted sample via a Y-shaped optical fiber.

4. The online component detection method based on the chromatographic elution process of Tanshinone extract as described in claim 1, characterized in that, The components to be tested include salvianolic acid A and salvianolic acid B.

5. The online component detection method based on the chromatographic elution process of Salvia miltiorrhiza extract as described in claim 4, characterized in that, Different fusion strategies include basic fusion strategies, intermediate fusion strategies, and advanced fusion strategies. The specific steps of a basic fusion strategy include: Near-infrared and ultraviolet-visible spectral data are normalized by mean and then spliced ​​in a certain order to form a new dataset containing the two spectral information, which is used to build the model. The specific steps of the intermediate fusion strategy include: Principal component analysis was used to extract features from near-infrared and ultraviolet-visible spectral data to obtain the near-infrared spectral feature matrix TUV-Vis and the ultraviolet-visible spectral feature matrix TNIR. Subsequently, the feature data were selectively fused according to the proportion of each principal component to construct and screen the best fusion model. The specific steps of the advanced integration strategy include: Advanced fusion builds upon primary fusion by constructing separate classification or prediction models using UV-Vis and NIR data. Then, based on the prediction results of different models, a decision-level fusion is performed using a stacking method to form the final fusion model.

6. The method for online component detection based on the chromatographic elution process of Tanshinone extract as described in claim 5, characterized in that, The content of salvianolic acid A was detected using a spectral fusion model constructed using an intermediate fusion strategy, and the content of salvianolic acid B was detected using a spectral fusion model constructed using an advanced fusion strategy.

7. The online component detection method based on the chromatographic elution process of Tanshinone extract as described in claim 1, characterized in that, The specific steps for processing near-infrared and ultraviolet-visible spectral data using different spectral fusion models are as follows: Preprocessing of near-infrared and ultraviolet-visible spectral data; The preprocessed near-infrared and ultraviolet-visible spectral data are converted into absorbance. The absorbance data were fused using a matched spectral fusion model.

8. An online component detection system based on the chromatographic elution process of Salvia miltiorrhiza extract, characterized in that, include: The experimental module is configured to perform elution experiments on the extract of Salvia miltiorrhiza based on column chromatography to obtain a sample set; The data acquisition module is configured to acquire near-infrared and ultraviolet-visible spectral data of the sample; The model matching module is configured to match spectral fusion models according to the type of the component to be detected, wherein different spectral fusion models are constructed based on different fusion strategies for near-infrared and ultraviolet-visible spectra. The online detection module is configured to process near-infrared and ultraviolet-visible spectral data using different spectral fusion models to obtain the content of the component to be detected.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program adapted to be loaded by a processor and executed as described in any one of claims 1-7, the method for online detection of components based on the chromatographic elution process of tanshinone extract.

10. A computer device, characterized in that, include: A processor, adapted to execute computer programs; A computer-readable storage medium storing a computer program, which, when executed by the processor, implements the online component detection method based on the chromatographic elution process of tanshinone extract as described in any one of claims 1-7.

Images