Design for optimizing honeysuckle rapid drying process based on multiple algorithms
Through the response surface method and genetic algorithm combined with BP neural network to optimize the honeysuckle drying process, the optimal drying parameters were determined, the problem of low drying efficiency of honeysuckle was solved, efficient and accurate drying process optimization was achieved, and the quality of medicinal materials and development prospects were improved.
Patent Information
- Application Number
- CN202410944277.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-15
- Publication Date
- 2024-11-12
AI Technical Summary
The drying process of honeysuckle has problems such as low efficiency and degradation of medicinal materials. The prior art is difficult to maintain the effective ingredient content of honeysuckle quickly and effectively, and the drying method is complex and there is a lack of efficient optimization methods.
Using response surface method and genetic algorithm combined with BP neural network, the honeysuckle drying process is optimized through various algorithms to determine the best drying parameters, including finishing time, drying temperature and drying time, and optimize the drying process to improve the comprehensive content of active ingredients. Score.
It has achieved rapid and accurate optimization of honeysuckle drying process, strong predictive model and small relative errors, providing a new drying process reference, and improving the quality of medicinal materials and development prospects.
Smart Images

Figure CN118940611A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the technical field of drying processes, and in particular to a design for optimizing a quick drying process for honeysuckle based on multiple algorithms. Background Art
[0002] Honeysuckle is a commonly used Chinese herbal medicine in my country. It is the dried buds or flowers of Lonicera Japonica Thunb., a plant of the Caprifoliaceae family. It can dispel wind-heat, clear away heat and detoxify. It is clinically used to treat wind-heat colds, sores, carbuncles, swellings and other diseases. Modern pharmacological studies have shown that honeysuckle has anti-inflammatory, antibacterial and antiviral activities.
[0003] Honeysuckle has the characteristics of short flowering period, high moisture content of fresh flowers, and rich enzyme types. Therefore, fresh flowers should be dried in time after harvesting to prevent browning and reduce the quality of medicinal materials. Common drying methods for honeysuckle include sun drying, shade drying, oven drying, and withering drying.
[0004] Response surface method is an effective method to reduce the number of experiments. It has a high model fitting efficiency and is therefore suitable for modeling and analysis of practical problems. BP (Back Propagation) neural network is one of the artificial neural network algorithms. Its advantages are that it can enhance the classification and recognition capabilities of neural networks and solve nonlinear problems. Since it can form a multi-layer feedforward perceptron network, it has better fitting effects and stronger prediction performance for highly complex functional relationships. However, the disadvantages of this algorithm are long training time, slow learning speed, and complex network structure. The advantages of genetic algorithm (GA) are strong global search capabilities, applicability to various problems, the ability to add prior knowledge to guide the search, and parallel implementation. However, the amount of calculation is large. In addition, the programming implementation of genetic algorithms is complex, parameter selection is difficult, and the problem needs to be decoded and encoded. Therefore, it is necessary to combine the advantages of the above three algorithms to make up for their respective shortcomings. Summary of the invention
[0005] The object of the present invention is to provide a design for optimizing the rapid drying process of honeysuckle based on multiple algorithms to solve the problems raised in the above background technology. The specific technical scheme is as follows:
[0006] A design for optimizing the rapid drying process of honeysuckle based on multiple algorithms includes the following steps:
[0007] ① Methodological investigation was conducted on four drying methods, namely shade drying, sun drying, oven drying, and green drying, and the drying effect was evaluated by the comprehensive score of the effective ingredient content, where the comprehensive score of the effective ingredient content was obtained by principal component analysis;
[0008] ② Experimental investigations were conducted on the single factors of fixing time, drying temperature and drying time, and the comprehensive scores of the effective ingredient content of each sample were obtained by principal component analysis, and the drying effect was evaluated by the comprehensive scores;
[0009] ③ The Box-Behnken method in response surface methodology was used to conduct a three-factor three-level experiment to obtain the optimal prediction value of drying process parameters;
[0010] ④ Apply response surface design analysis to establish a BP neural network model, and test and train the model with different numbers of hidden layer neurons;
[0011] ⑤Use genetic algorithm to optimize BP model and conduct target optimization of honeysuckle drying process parameters;
[0012] ⑥ Conduct verification tests on the found optimization conditions and calculate the relative error between the experimental measurement value and the predicted value.
[0013] Furthermore, in step ①, the better drying method obtained through methodological investigation is withering drying.
[0014] Furthermore, in step ②, the central point of the fixing time of the response surface is selected as 2 min, the central point of the drying temperature is selected as 50° C., and the central point of the drying time is selected as 5 h.
[0015] Furthermore, in step ③, the factors affecting the fixing and drying process are fixing time>drying time>drying temperature.
[0016] Furthermore, in step ④, there are 3 input nodes, namely, the fixing time, the drying temperature, and the drying time; there is 1 output node, which is the comprehensive evaluation value; the fitting error and the prediction error are used as the basis for debugging and training the parameters of the neural network model, and the training results are obtained, and finally the number of hidden layer neurons is selected as 9.
[0017] Furthermore, the results obtained after optimization by genetic algorithm are as follows: after 100 iterations, the fitness is high and stable, and the predicted comprehensive score is 47.3072; the optimal conditions are: fixing time is 1.94635min, drying temperature is 50.3691℃, and drying time is 5.78133h; based on the actual situation, the above process parameters are revised to: fixing time 2min, drying temperature 50℃, and drying time 6h.
[0018] Compared with the prior art, the beneficial effects of the present invention are as follows: the present invention adopts BP neural network combined with genetic algorithm under MATLAB language environment to optimize the process and seek the target for BBD experimental results, the model is fast and accurate, and provides a new reference and basis for the optimization of honeysuckle drying process; the optimized extraction process is verified, and the relative error between the experimental measurement value and the network prediction value is 1.8%<5%, indicating that the model has good predictability; honeysuckle is widely used, and is used for wind-heat cold, sore, carbuncle, swelling and other diseases because of clearing away heat and detoxifying, and evacuating wind-heat, and has good development prospects. BRIEF DESCRIPTION OF THE DRAWINGS
[0019] Figure 1 It is a high performance liquid chromatogram for determining the content of the active ingredient of the present invention;
[0020] Figure 2 The effect of the fixing time of the present invention on the comprehensive score of the content of the active ingredients of honeysuckle;
[0021] Figure 3 The effect of the drying temperature of the present invention on the comprehensive score of the effective ingredient content of honeysuckle;
[0022] Figure 4 The effect of the drying time of the present invention on the comprehensive score of the effective ingredient content of honeysuckle;
[0023] Figure 5 is the positive probability distribution of the response surface design defect of the present invention;
[0024] Figure 6 Comparison between the actual and predicted values of the response surface design of the present invention;
[0025] Figure 7 Comparison of residuals and operations for the response surface design of the present invention;
[0026] Figure 8 Cook's distance for the response surface design of the present invention;
[0027] Fig. 9 The response surface of the fixing time and the drying time for the comprehensive score of the effective component content of honeysuckle in the present invention;
[0028] Fig.10 The response surface of the fixing time and the drying temperature on the comprehensive score of the effective component content of honeysuckle in the present invention;
[0029] Fig.11 The response surface of the drying time and the drying temperature to the comprehensive score of the effective component content of honeysuckle in the present invention;
[0030] Fig.12 is the MSE value of the present invention with different numbers of hidden layer neurons;
[0031] Fig.13 It is the BP model training diagram of the present invention;
[0032] Fig.14 This is the training state diagram of the BP model of the present invention;
[0033] Fig.15 is the error histogram of the BP model of the present invention;
[0034] Fig.16 It is the regression diagram of the BP model of the present invention;
[0035] Fig.17 This is the result of GA optimizing the BP model of the present invention. DETAILED DESCRIPTION
[0036] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0037] A methodological investigation was conducted on four drying methods of honeysuckle, including shade drying, sun drying, oven drying, and withering drying. The drying effect was evaluated by the comprehensive score of the effective ingredient content (obtained by principal component analysis). The results of the methodological investigation showed that the samples obtained by withering drying had higher drying quality.
[0038] The experimental process is as follows:
[0039] The main instruments and software used in the experiment are: Dionex UltiMate3000 high performance liquid chromatograph (ThermoFisher, USA); DHG-9146A electric constant temperature blast drying oven (Shanghai Jinghong Experimental Equipment Co., Ltd.); G80F20CN2L-B8 (R0) microwave oven (Foshan Galanz Microwave Oven Appliance Co., Ltd.); pipettes, conical flasks and other glass instruments; Origin Pro 2021, Excel 2019, SPSS 27.0, Design-Expert 13, MatLab R2021a software;
[0040] The reagents and medicinal materials used in the experiment were: honeysuckle (purchased on May 13, 2024 at the Chunnuan Honeysuckle Professional Cooperative in Julu County, Xingtai City, Hebei Province (115°04′E, 37°19′N), identified as Lonicera japonica of the Caprifoliaceae family buds of Thunb.); chlorogenic acid (CAS: 327-97-9) was purchased from Chengdu Pusi Biotechnology Co., Ltd.; 3,5-O-dicaffeoylquinic acid (CAS#89919-62-0), 4,5-di-O-caffeoylquinic acid (CAS#57378-72-0), and luteolin (CAS#5373-11-5) were purchased from Shanghai Yuanye Biotechnology Co., Ltd., with a mass fraction of ≥98%; methanol and anhydrous ethanol (analytical grade; Tianjin Fuyu Fine Chemical Co., Ltd.); phosphoric acid (chromatographic grade; Tianjin Yongda Chemical Reagent Development Center); methanol and acetonitrile (chromatographic grade; Tianjin Fuchen Chemical Reagent Co., Ltd.); Wahaha purified water (Hangzhou Wahaha Group Co., Ltd.).
[0041] The specific experimental methods include the following:
[0042] Drying method 1: Natural drying in the shade (YG): Spread the fresh honeysuckle flat and place it in a cool, dry and ventilated place. Observe the state of the medicinal material every day until it is completely dried, which takes about 6-7 days.
[0043] Method 2, natural sun drying (SG): Spread the fresh honeysuckle flat and place it in a sunny place with sufficient sunshine. Dry it on the same day and collect it on the same day. Observe the condition of the medicinal material every day until it is completely dried. It takes about 2-3 days.
[0044] Method 3, constant temperature hot air drying (R1-R3): Spread the fresh honeysuckle on a tray, set the temperature of the blast drying oven, put the herbs into the blast drying oven for constant temperature drying after the temperature reaches the set value, and take out the herbs after the set time is reached;
[0045] Method 4, withering and drying (S1-S9): put fresh honeysuckle on a steamer, steam and wither, take out, spread it evenly on filter paper, and put it into a blast drying oven for constant temperature drying.
[0046] The active ingredients were determined by the following method:
[0047] Chromatographic conditions: The chromatographic column was a Sharpsil-AR ODS column (250 mm × 4.6 mm, 5 μm), with 0.1% phosphoric acid water (A)-acetonitrile (B) as the mobile phase, and gradient elution: 0-5 min, 85%-80% A; 5-7 min, 80%-80% A; 7-22 min, 80%-70% A; 22-30 min, 70%-65% A; 30-34 min, 65%-65% A; 34-40 min, 65%-60% A; 40-45 min, 60%-40% A; 45-50 min, 40%-85% A; 50-55 min, 85%-85% A. The flow rate was 0.8 mL / min; the detection wavelength was 327 nm; the column temperature was 25°C; the injection volume was 10 μL, and the chromatogram of the reference substance is shown in Figure 1 (Peaks 1, 2, 3, and 4 in the figure are chlorogenic acid, luteolin, 3,5-O-dicaffeoylquinic acid, and 4,5-O-dicaffeoylquinic acid, respectively);
[0048] Preparation of reference solution: Accurately weigh four reference substances and prepare reference solutions with concentrations of 0.102 40 mg / mL chlorogenic acid, 0.106 68 mg / mL luteolin, 0.10012 mg / mL 3,5-O-dicaffeoylquinic acid, and 0.118 00 mg / mL 4,5-O-dicaffeoylquinic acid using 75% methanol. Use a manual pipette to draw 200 μL of each of the four reference solutions to prepare a reference mixed solution;
[0049] Preparation of test solution: crush honeysuckle and pass through No. 4 sieve. Accurately weigh 2.0g honeysuckle powder, place in a 100mL conical flask, add 50mL of 75% methanol, weigh, ultrasonically extract for 40min, cool to room temperature, weigh again and add 75% methanol, shake well, take the supernatant and filter through a 0.45μm microporous membrane, place in a liquid phase vial, and obtain;
[0050] Calculation of active ingredient content: Take the reference solution of 4 index components and prepare 5 series of reference solution. Inject and measure according to the chromatographic conditions to obtain the standard curve equations of the 4 components. Determine the peak area of each batch of samples according to the standard curve drawing method, and calculate the content of each active ingredient in the test solution according to the standard curve.
[0051] Calculation of comprehensive score: The principal component analysis was performed with the contents of the four active ingredients and phenolic acids as variables, and the calculation formula for the comprehensive score was obtained:
[0052] Y1=0.513X1+0.512X2+0.5X3+0.427X4+0.206X5,
[0053] Y2=-0.032X1-0.073X2+0.021X3-0.348X4+0.935X5,
[0054] Y=0.7422X1+0.1868X2.
[0055] Each sample was scored, and the ranking of effective ingredient content and score is shown in Table 1. As can be seen from Table 1, the samples obtained by drying and fixing generally scored higher, among which the sample obtained by fixing for 2 minutes and then drying at 50°C for 5 hours had the highest comprehensive score.
[0056] Table 1 Comprehensive scores and rankings of principal components of each sample
[0057]
[0058]
[0059] Single factor experimental design content: to investigate the effect of the killing time (1, 2, 3, 4, 5min) on the comprehensive score of the effective ingredient content of honeysuckle. Figure 2 It can be seen that 2 min is the appropriate central point of the killing time of the response surface;
[0060] The effect of drying temperature (30, 40, 50, 60, 70°C) on the comprehensive score of the active ingredient content of honeysuckle was investigated. Figure 3 It can be seen that 50°C is the appropriate center point of the drying temperature on the response surface;
[0061] The effect of drying time (3, 4, 5, 6, 7 h) on the comprehensive score of the active ingredient content of honeysuckle was investigated. Figure 4 It can be seen that 5h is the appropriate central point of the drying time in the response surface;
[0062] Based on the experimental principle (Box-Behnken Design, BBD) in Design expert 13 software. According to the results of single factor, the design level table of response surface is as follows:
[0063] Table 2 Process optimization factor level table
[0064]
[0065] Response surface design results and analysis Figure 5-8 As shown in Tables 3 and 4.
[0066] Table 3 Response surface experimental design results
[0067]
[0068] Table 4 Analysis of variance of regression model
[0069]
[0070]
[0071] Establishing regression equation: 17 groups of response surface designs and their experimental results are shown in Table 3 above. The above data were analyzed by response surface analysis using DesignExpert.13 software, and the multiple regression equation of the three factors affecting the drying of honeysuckle and the response surface was obtained as follows:
[0072] Y=40.02-0.7864A-0.2085B+0.6749C+1.34AB+1.61AC+1.27BC-2.67A2-1.39B2-3.92C 2
[0073] Equation analysis: ANOVA see Table 4, combined with Figure 5-8 It can be seen that the P value of the model is less than 0.05, the model is significant, and the P value of the lack of fit term is greater than 0.05, which is not significant, indicating that the fitting equation has a good regression effect; R 2 =0.841 4, R 2 Adj =837 6, CV = 5.46%, indicating that the fitting equation has high credibility and can be used to analyze and predict the process of honeysuckle drying.
[0074] Corresponding surface analysis:
[0075] Depend on Figure 9-11 From the data processing results, it can be seen that the optimal drying process prediction values are fixing time 1.872min, drying temperature 59.759℃, and drying time 5.972h; considering the experimental conditions, operability and rationality, the factors are set as fixing time (A): 2min, drying temperature (B) 60℃, and drying time (C) 6h.
[0076] Establishment of BP neural network model:
[0077] BP neural network was used for modeling. The number of input nodes was 3, namely, killing time, drying temperature, and drying time. The number of output nodes was 1, which was the comprehensive evaluation value. The fitting error and prediction error were used as the basis for training the parameters of the neural network model, and other parameters were set to default values. 17 groups of central combination design experimental data were used as training samples, and the model was tested and trained for different numbers of hidden layer neurons. The MSE value was used as the basis for selecting the number of hidden layer neurons. The training results are shown in Fig.12 , and finally the number of hidden layer neurons is selected as 9.
[0078] BP training results are as follows Figure 13-16 described.
[0079] like Fig.13As shown in Figure 2, during the training, validation, and testing process, as the number of weight update iterations increases, the curve changes from underfitting to optimal fitting. When the validation error reaches the 11th generation, the gradient value and Valfail value are 1.08×10 -2 and 1, indicating that this BP model has been very intensively trained and the model training is stopped; Fig.14 As shown in , the best validation performance is obtained after 10 iterations, with an MSE of 0.0570; Fig.15 As shown in , the maximum fitting error between the BP model prediction value and the experimental value is 0.0163, which is very close to zero, indicating that the results predicted by this model are reliable; Fig.16 As shown, the R values in the training, validation, test, and prediction graphs are all greater than 0.9, indicating that the predicted values of the model match the actual values well and can well predict the test results.
[0080] Genetic algorithm optimization:
[0081] Matlab simulation was used to simulate the natural evolution process to obtain the optimal solution. The fitness curve obtained by GA operation is shown in Fig.17 ;like Fig.17 As shown in the figure, after 100 iterations, the fitness is the highest and stable, Y reaches the maximum value, and GA stops running. At this time, the level encoding of X1, X2, and X3 is brought into the constraints optimized by GA, and the levels of each factor are 1.946min, 50.369℃, and 5.781h. Under this condition, the maximum comprehensive score is 47.3068; considering the experimental conditions, operability and rationality, the conditions are approximated as: 2min of fixing time, 50℃ of drying temperature, and 6h of drying time.
[0082] Predicted process parameter validation:
[0083] The predicted process parameters of the two optimization methods were verified three times, and the results are shown in Table 5.
[0084] Table 5 Drying process verification results (n=3)
[0085]
[0086] As shown in Table 5, the comprehensive score of the process prediction value obtained by GA optimization on the BP model is higher than that of the response surface optimization, and the RSD value is smaller (P < 0.05). Therefore, the optimal drying process conditions finally obtained are: fixing time 2min, drying temperature 50℃, and drying time 6h.
[0087] Single factor analysis: Through single experiments, we roughly found the relatively optimal intervals, among which the best single factor was 2min, 50℃, and 5h. Honeysuckle has good antibacterial and antiviral effects. At the same time, honeysuckle has complex and multi-effect ingredients, which has good development significance. Therefore, the comprehensive score of the effective ingredient content was selected as an indicator for comprehensive evaluation in the single factor experiment.
[0088] Response surface analysis: This experiment determined the horizontal center point of the response surface as 2min, 50℃, 5h through a single factor experiment. The P value of the model is less than 0.05, which is significant, and the P value of the lack of fit term is greater than 0.05, which is not significant, indicating that the fitting equation has a good regression effect.
[0089] BP neural network modeling: This experiment uses BP neural network in the MATLAB language environment to optimize the BBD test results. There are few reports on its application in the drying process of Chinese herbal medicines. This model is fast and accurate, and provides a new reference and basis for the optimization of honeysuckle drying process. In this experiment, when the number of hidden layer neurons is 9, the MSE is the smallest, that is, 3-9-1 is the best topology for optimizing process conditions. The maximum fitting error between the predicted value and the experimental value of the BP model is 0.0163, which is very close to zero, indicating that this model is reliable.
[0090] GA optimization: Using genetic algorithm to optimize based on BP neural network can make the prediction results more accurate. Matlab simulation is used to simulate the natural evolution process to obtain the optimal solution. After 100 iterations, the fitness is the highest and stable, Y reaches the maximum value, and GA stops running.
[0091] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.
Claims
1. A design for optimizing the rapid drying process of honeysuckle based on multiple algorithms, characterized in that: The steps include: ① Methodological investigation was conducted on four drying methods, namely shade drying, sun drying, oven drying, and green drying, and the drying effect was evaluated by the comprehensive score of the effective ingredient content, where the comprehensive score of the effective ingredient content was obtained by principal component analysis; ② Experimental investigations were conducted on the single factors of fixing time, drying temperature and drying time, and the comprehensive scores of the effective ingredient content of each sample were obtained by principal component analysis, and the drying effect was evaluated by the comprehensive scores; ③ The Box-Behnken method in response surface methodology was used to conduct a three-factor three-level experiment to obtain the optimal prediction value of drying process parameters; ④ Apply response surface design analysis to establish a BP neural network model, and test and train the model with different numbers of hidden layer neurons; ⑤Use genetic algorithm to optimize BP model and conduct target optimization of honeysuckle drying process parameters; ⑥ Conduct verification tests on the found optimization conditions and calculate the relative error between the experimental measurement value and the predicted value.
2. The method according to claim 1, wherein the method is characterized in that: In step ①, the best drying method obtained from methodological investigation is withering drying.
3. The method according to claim 1, wherein the method is characterized in that: In step ②, the central point of the fixing time of the response surface is selected as 2 min, the central point of the drying temperature is selected as 50°C, and the central point of the drying time is selected as 5 h.
4. The method according to claim 1, wherein the method is characterized in that: In step ③, the factors affecting the fixing and drying process are fixing time>drying time>drying temperature.
5. The method according to claim 1, wherein the method is characterized in that: In step ④, there are 3 input nodes, namely, the fixing time, the drying temperature, and the drying time; there is 1 output node, which is the comprehensive evaluation value; the fitting error and the prediction error are used as the basis for debugging the parameters of the training neural network model to obtain the training results, and finally the number of hidden layer neurons is selected as 9.
6. A design of a rapid process for optimizing Flos Lonicerae based on multiple algorithms according to claim 1, characterized in that: The results obtained after optimization by genetic algorithm are as follows: after 100 iterations, the fitness is high and stable, and the predicted comprehensive score is 47.3072; the optimal conditions are: fixing time is 1.94635min, drying temperature is 50.3691℃, and drying time is 5.78133h; based on the actual situation, the above process parameters are revised to: fixing time 2min, drying temperature 50℃, and drying time 6h.
Citation Information
Cited By
Intelligent optimization method of wine rhizoma polygonati processing technology
CN120544725A
