CN104865322A - Rapid detection method for concentration process of Fructus Gardeniae extract liquor - Google Patents

Abstract

The invention provides a rapid detection method for the concentration process of the gardenia extract, which measures the key indicators in the gardenia extract concentrate by collecting different batches of the gardenia extract concentrate samples in the large-scale production of Reduning injection, including To determine the solid content and concentration of geniposide, collect the near-infrared spectrum of the concentrated liquid sample of Gardenia jasminoides extract, select the appropriate spectral preprocessing method and modeling band, and use the partial least square method and forward artificial neural network method respectively to establish The near-infrared quantitative calibration model of solid content and geniposide concentration was used to examine the performance of the model, and the verification set data was imported into the established calibration model to evaluate the predictive ability of the model. The present invention introduces near-infrared spectroscopy technology as a detection method for the concentration process of the gardenia extract, which can quickly detect the change of the solid content and the concentration of geniposide in the concentration process of the gardenia extract, and the established analysis method meets the requirements of rapid and effective actual production requirements and has broad application prospects.

Description

A rapid detection method for the concentration process of gardenia extract

technical field

The invention belongs to the field of near-infrared rapid detection, and in particular relates to a rapid detection method for the concentration process of gardenia extract.

Background technique

The main components of Reduning Injection are honeysuckle, Artemisia annua and Gardenia, and it has significant curative effect on influenza, acute upper respiratory tract infection, acute bronchitis, and respiratory medicine. The concentration process after extraction is one of the key processes in the production of Reduning, which is directly related to the change of the content of each component in Reduning. At present, the quality control of the post-extraction concentration process mainly relies on experience and traditional quality analysis methods, which is time-consuming and laborious. Therefore, the research and development of rapid and non-destructive determination methods for key quality control indicators in the concentration process of Reduning gardenia extract will help to solve thermal problems. The quality control of key control indicators in the production process of Duning has great practical significance for the technological progress and product quality upgrading of the traditional Chinese medicine industry.

As an important representative of process analysis technology, the near-infrared spectroscopy technology that has emerged in recent years has the advantages of fast detection speed, simple sample processing and no loss, etc., and has broad application prospects. The application of near-infrared spectroscopy technology in petroleum, tobacco, pharmaceuticals and other fields has also made good progress. Especially in recent decades, with the continuous development of chemometrics, the application of near-infrared spectroscopy in process analysis has become more and more extensive.

Contents of the invention

The object of the present invention is to provide a kind of rapid detection method of gardenia extract concentration process. The model established by this method can quickly and accurately determine the solid content and geniposide concentration in the concentrated solution of gardenia extract in the production of Reduning injection.

The purpose of the present invention is achieved through the following technical solutions:

1. Collect samples of different batches of gardenia extract concentration process in the large-scale production of Reduning injection

During the concentration process of gardenia extract, the temperature is controlled to be ≦70°C, the volume of each batch of liquid medicine is 500~600kg, and about 2 tons of n-butanol is recovered in each batch, and it takes 8~9 hours to take samples at intervals of 20~40 minutes. ~10 min sampling, each sampling about 10 mL, 16~38 samples per batch. 6 batches, 127 samples. All samples were collected from the same sampling port of the single-effect concentration tank.

2. Determination of key indicators

The concentration and solid content of geniposide in the samples during the concentration process of gardenia extract were determined by high performance liquid chromatography and drying weighing method.

(1) Determination of geniposide concentration in the concentrated sample of gardenia extract by high performance liquid chromatography:

Precisely draw 1 mL of the concentrated sample of gardenia extract into a 100 mL measuring bottle, add 50% methanol to dissolve, and dilute to the mark, shake well, then precisely draw 1 mL into a 10 mL measuring bottle, add 50% methanol to dissolve , and dilute to the mark, shake well and filter the resulting liquid with a 0.45μm organic filter membrane, which is the analytical sample injection. Liquid chromatography conditions: chromatographic column: Kromasil C18 (4.6×250 mm, 5μm); mobile phase: acetonitrile (A)-0.2% glacial acetic acid solution (B), according to acetonitrile: 0.2% glacial acetic acid=12:88 (v: v) ratio isocratic elution; detection wavelength: 237 nm; flow rate: 1 mL/min; column temperature: 30 °C.

(2) Determine the solid content in the concentrated sample of gardenia extract by drying and weighing method:

Let the concentrated sample of gardenia extract stand for 24 hours, measure 3 mL of supernatant to a flat bottle that has been dried to constant weight (the difference in weight after two dryings is less than 5 mg, count as X ₀ ), weigh ( X ₁ ), dried in an oven at 105°C until the weight difference between the two weighings is less than 5 mg, count as X ₂ .

Then solid content (%)=(X ₂ -X ₀ )/(X ₁ -X ₀ )×100%.

3. Near-infrared spectroscopy data collection

Heat the concentrated sample of gardenia extract to 70°C in a water bath, and then accurately measure 1ml of the concentrated sample of gardenia extract, place it in a U-shaped small tube matched with a Nicolet ANTARISII near-infrared spectrometer, and collect near-infrared spectra by transmission method with 32 scans. The resolution is 4 cm ^-1 , the optical path of the optical fiber transmission probe is 2 mm, and the scanning spectrum ranges from 4500 to 12000 cm ^-1 with air as the reference.

4. The partial least squares method and the forward artificial neural network method were used to establish the near-infrared quantitative model of each quality control index in the concentration process of gardenia extract. When the correlation coefficient value is close to 1, the root mean square error of the calibration set and the validation set When the root mean square error is small and close to each other, it indicates that the model has high predictive ability and good stability. In addition, when the relative prediction deviation value is less than 10%, the model is considered to have good predictive ability and can be used for quantitative control of indicators.

The Result TQ Analyst 8.0 data processing software provided by Nicolet ANTARISII near-infrared spectrometer was used. Before establishing the calibration model, due to factors such as baseline drift and noise interference, the spectrum was firstly preprocessed and the band was selected. The partial least squares method (PLS) and forward artificial neural network method (BP-ANN) were used to establish the quantitative calibration model of the two quality control indicators of near-infrared data and solid content, near-infrared data and geniposide concentration, and passed Each model evaluation index examines the model performance. Import the verification set data into the established calibration model, and judge the stability and predictive ability of the model through the model performance evaluation index.

Specifically, in the above step (4), the preprocessing method of the geniposide concentration quantitative correction model selects the second derivative and the Savitsky-Golay filter smoothing preprocessing, which are respectively used to eliminate baseline drift and noise. According to the correlation between wavelength and spectral information, the quantitative correction model of geniposide concentration was finally established by using the 6013~4793 cm ^-1 band; the preprocessing method of the solid content model included first-order derivative and Norris smoothing, and the band was selected as 5769 ^~ 5677 cm -1 ¹ and 5893~5831 cm ^-1 . Quantitative correction model established by partial least squares method and forward artificial neural network method. Model evaluation indicators include: correlation coefficient (R), root mean square error of correction set (RMSEC), root mean square error of verification set (RMSEP) and external The main parameter of validation was relative prediction error (RSEP). When the correlation coefficient value is close to 1, the root mean square error of the calibration set and the root mean square error of the validation set are small and close to each other, indicating that the model has high predictive ability and good stability. In addition, when the relative prediction deviation value is less than 10%, the model is considered to have good predictive ability and can be used for quantitative control of indicators.

Another object of the present invention is to provide the application of the method in the rapid detection of the gardenia extract concentration process.

The invention combines the near-infrared spectrum technology with the partial least square method and the forward artificial neural network algorithm, and applies it to the determination of key indicators in the process of concentrating the gardenia extract in Reduning injection. The method is simple to implement, has strong model prediction ability, high stability, and strong extrapolation and generalization capabilities.

Description of drawings

Fig. 1 is the change trend of the concentration of geniposide during the concentration process of the gardenia extract.

Fig. 2 is the variation trend of the solid content in the process of concentrating the gardenia extract.

Figure 3 is the original near-infrared spectrum during the concentration process of the gardenia extract.

Fig. 4 is a graph of correlation coefficients in each band of the near-infrared spectrum.

Figure 5 is a comparison chart of the predicted value of the concentration of geniposide in the concentration process of the gardenia extract obtained by using the partial least squares method and the actual measurement trend.

Figure 6 is a comparison chart of the predicted value of the solid content in the extraction and concentration process of Gardenia jasminoides obtained by modeling using the partial least squares method and the actual measurement trend.

Figure 7 is a comparison chart of the predicted value of geniposide concentration in the concentration process of the gardenia extract obtained by using the forward artificial neural network method and the actual measurement trend.

Figure 8 is a comparison chart of the predicted value of the solid content in the concentration process of the gardenia extract obtained by using the forward artificial neural network method and the actual measurement trend.

Detailed ways

The present invention will be further described in conjunction with drawings and embodiments.

Example 1: Establishment of a near-infrared quantitative analysis model for solid content and geniposide concentration using the partial least squares method

1. Collection of samples of gardenia extract concentrate

During the concentration process of gardenia extract, the temperature is controlled to be ≦70°C, the volume of each batch of liquid medicine is 500~600kg, and about 2 tons of n-butanol is recovered in each batch, and it takes 8~9 hours to take samples at intervals of 20~40 minutes. ~10 min sampling, each sampling about 10 mL, 16~38 samples per batch. 6 batches, 127 samples. All samples were collected from the same sampling port of the single-effect concentration tank.

2. Determination of key indicators

(1) Determination of geniposide concentration

Chromatographic conditions: chromatographic column: Kromasil C18 (4.6×250 mm, 5 μm); mobile phase: acetonitrile (A) - 0.2% glacial acetic acid solution (B), isocratic washing according to the ratio of acetonitrile: 0.2% glacial acetic acid = 12:88 detection wavelength: 237 nm; flow rate: 1 mL/min; column temperature: 30 ℃.

Precisely draw 1 mL of the concentrated sample into a 100 mL volumetric flask, add 50% methanol to dissolve, dilute to the mark, and shake well. Then accurately draw 1 mL to a 10 mL measuring bottle, add 50% methanol to dissolve, and dilute to the mark, shake well, and filter through a 0.45 μm organic filter membrane to obtain the analytical sample.

The change trend of geniposide concentration in the concentration process of gardenia extract is shown in Figure 1.

(2) Determination of solid content

The concentrated solution of gardenia extract was left to stand for 24 hours, and 3 mL of the supernatant was taken to a flat bottle that had been dried to a constant weight (the weight difference between two times of drying was less than 5 mg, which was counted as X ₀ ), and weighed (X ₁ ) Dry in an oven at 105°C until the weight difference between the two weighings is less than 5 mg, count as X ₂ . Then solid content (%)=(X ₂ -X ₀ )/(X ₁ -X ₀ )×100%.

The change trend of the solid content in the concentration process of the gardenia extract is shown in Figure 2.

3. Near-infrared spectroscopy data collection

Use ANTARIS Fourier transform near-infrared spectrometer (Thermo Nicolet, USA) to collect the near-infrared transmission spectrum of gardenia extract concentrate sample, the band range is 4500~12000 cm ^-1 , the number of scans is 32 times, ^and the resolution is 4 cm -1 ¹ . Take air as a reference.

The original near-infrared spectrum collected during the concentration process of the gardenia extract is shown in Figure 3.

4. Establishment of partial least squares quantitative model

Modeling Band Selection and Spectral Preprocessing

Since gardenia extract is concentrated to recover n-butanol, which contains hydroxyl groups and has strong polarity, combined with the correlation coefficient diagram of each band of near-infrared spectrum, the optimal band 6013~4793 cm ^-1 is finally selected as the modeling band for the quantitative model of geniposide , the solid content model selects the 5769~5677 cm ^-1 and 5893~5831 cm ^-1 bands with a spectral correlation coefficient greater than 0.8 and relatively concentrated information. The correlation coefficient diagram of each band of near-infrared spectrum is shown in Fig. 4.

In order to effectively eliminate the baseline shift, reduce the overlap between peaks and make effective information appear, the derivative processing method is often used, and in order to reduce the noise introduced by the derivative calculation and reduce the signal-to-noise ratio, the smoothing method is used to reduce high-frequency random noise . Therefore, for the establishment of the geniposide concentration model, the present invention uses the second-order derivative method and S-G smoothing filter to preprocess the spectral data; for the correction model of the solid content, the first-order derivative and Norris smoothing are used for preprocessing. Randomly select 14 samples as the verification set prediction, and the remaining 113 samples as the calibration set

Table 1 shows the parameters of the near-infrared model modeling results of geniposide and solid content. It can be seen from Table 1 that the near-infrared models established by the partial least squares method of geniposide and solid content have good linearity, the correlation coefficients are all greater than 0.90, and the root mean square error RMSEC of the calibration set is close to the root mean square error RMSEP of the verification set It shows that the established near-infrared quantitative model has better effect, better predictive ability and better robustness.

5. Validation of the partial least multiplication quantitative correction model

The near-infrared quantitative calibration model established by the partial least squares method was used to predict the geniposide concentration and solid content of the 14 validation samples in the corresponding model. The comparison between the measured value of geniposide concentration and the near-infrared predicted value is shown in Figure 5, and the comparison between the measured value of solid content and the predicted value of near-infrared is shown in Figure 6. near. The RSEP values of the geniposide and solid content quantitative models were 7.09% and 5.59%, respectively, both of which were controlled within 8%, meeting the actual production accuracy requirements, indicating that the quantitative calibration model established by the partial least squares method can be used for the concentration of gardenia extract The process of geniposide concentration and solid content prediction, and to achieve a certain accuracy of prediction.

Example 2: Establishing a near-infrared quantitative analysis model for solid content and geniposide concentration using the forward artificial neural network method

1. Establishment of quantitative model of forward artificial neural network method

As mentioned above, after completing the determination of key indicators and spectral scanning of the concentrated samples of gardenia extract, 16 samples were randomly selected as the verification set prediction, and the remaining 111 samples were used as the calibration set. For the geniposide concentration and solid content, specific bands were selected (The quantitative model of geniposide concentration is in the band of 6013~4793 cm ^-1 , the quantitative model of solid content is in the band of 5769~5677 cm ^-1 and 5893~5831 cm ^-1 ). After processing (the geniposide concentration quantitative correction model is the combination of the second derivative method and SG smoothing filter, and the solid content quantitative correction model is the first derivative and Norris smoothing), the forward artificial neural network method is used to establish the geniposide and solid content respectively. Near-infrared quantitative calibration model.

Table 2 shows the parameters of the near-infrared model modeling results of geniposide concentration and solid content. It can be seen from Table 2 that the near-infrared models of geniposide concentration and solid content established by the forward artificial neural network method have good linearity, and the correlation coefficients are all greater than 0.90. RMSEP is small and very close to each other, indicating that the established near-infrared quantitative model has good prediction effect and model stability.

1. Verification of the quantitative model of the forward artificial neural network method

The near-infrared quantitative calibration model established by the forward artificial neural network method was used to predict the geniposide concentration and solid content of the 16 validation samples in the corresponding model. The comparison between the measured value of geniposide concentration and the near-infrared predicted value is shown in Figure 7, and the comparison between the measured value of solid content and the predicted value of near-infrared is shown in Figure 8. It can be seen that the measured value of geniposide and solid content is very close to the predicted value of near-infrared near. The RSEP values of the geniposide concentration and solid content quantitative models were 9.03% and 6.68%, respectively, and the RSEPs were less than 10%, meeting the actual production accuracy requirements, indicating that the quantitative calibration model established by the artificial neural network method can be used for the concentration of gardenia extract The process of geniposide concentration and solid content prediction, and to achieve a certain accuracy of prediction.

From the modeling parameters obtained in Table 1 and Table 2 above, it can be seen that under the condition of selecting the same preprocessing and modeling bands, the correlation coefficients of the model correction sets obtained by the partial least squares method and the forward artificial neural network algorithm are both greater than 0.9; In the geniposide quantitative model and the solid content quantitative model obtained by the algorithm, the root mean square error of the calibration set and the root mean square error of the verification set are relatively close; for the relative prediction deviation value, although both are controlled within 10%, but by the minimum deviation The geniposide and solid content models established by the square method were correspondingly better than those established by the forward artificial neural network method. It shows that the partial least squares model has a better predictive ability in the near-infrared rapid detection of the concentration process of gardenia extract.

Claims (6)

1. a cape jasmine extract concentration process method for quick, be is characterized in that, realized by following steps:

(1) the cape jasmine extract concentration process sample of different batches in the large production of Reduning injection is gathered;

(2) solid content and the Gardenoside concentration of cape jasmine extract concentration process sample is measured;

(3) the NIR transmittance spectroscopy figure of cape jasmine extract concentration process sample is gathered;

(4) partial least square method and Forward Artificial Neural Network method is adopted to set up the near infrared quantitative model of each quality control index in cape jasmine extract concentration process respectively; When correlation coefficient value is close to 1, calibration set error mean square root and checking collection error mean square root less and close to each other time, illustrate that the predictive ability of model is higher, good stability, think that model has good predictive ability when relative prediction deviation value is less than 10%, can be used for the fixing quantity of index.

2. require described a kind of cape jasmine extract concentration process method for quick according to right 1, it is characterized in that: step (1) is described in the large production of Reduning injection, Wen Du≤70 DEG C processed are controlled in cape jasmine extract concentration process, often criticizing amount of liquid medicine is 500 ~ 600kg, often criticizes and reclaims normal butyl alcohol about 2 tons, 8 ~ 9 hours used times, interval 20 ~ 40 min samples, last 1 h apart 5 ~ 10 min samples, every sub-sampling about 10 mL, and all sample standard deviations are from the same sample tap collection of single-action concentration tank.

3. require described a kind of cape jasmine extract concentration process method for quick according to right 1, it is characterized in that: the preprocess method of the Fructus Gardeniae by HPLC glycosides concentration described in step (2) is: concentrating sample precision is drawn in 1 mL to 100 mL measuring bottle, add 50% methyl alcohol to dissolve, and be diluted to scale, shake up, again in accurate absorption 1 mL to 10 mL measuring bottle, add 50% methyl alcohol to dissolve, and be diluted to scale, shake up, filter with 0.45 μm of organic filter membrane, to obtain final product; Weighting method after dried described in step (2) measures solid content method: after cape jasmine extract concentrate leaves standstill 24 h, measures 3 mL supernatants to the flat bottle (X of drying to constant weight ₀), be weighed as (X after adding cape jasmine extract concentrate ₁), drying under putting baking oven 105 DEG C of conditions, to twice weight difference of weighing apart from being less than 5 mg, counting X ₂, solid content computing formula is (%)=(X ₂-X ₀)/(X ₁-X ₀) × 100%.

4. require described a kind of cape jasmine extract concentration process method for quick according to right 1, it is characterized in that: after cape jasmine extract being concentrated in step (3) sample heating water bath to 70 DEG C, precision measures in the supporting U-shaped tubule of 1ml to Nicolet ANTARISII near infrared spectrometer, transmission beam method is adopted to gather near infrared spectrum, scanning times is 32, and resolution is 4 cm ^-1, after repeating the scanning of dress sample by 3 times, be averaging the final spectrogram obtaining this sample.

5. require described a kind of cape jasmine extract concentration process method for quick according to right 1, it is characterized in that: in step (4), Pretreated spectra condition is that Gardenoside concentration model adopts second derivative and S-G filtering Pretreated spectra; Solid content model adopts the level and smooth Pretreated spectra of first order derivative+Norris; Gardenoside model modeling wave band is 6013 ~ 4793 cm ^-1, solid content model modeling wave band is 5769 ~ 5677 cm ^-1with 5893 ~ 5831 cm ^-1.

6. method described in claim 1 is applied in cape jasmine extract concentration process detects fast.