Rapid detection method for moisture in low-cost traditional Chinese medicine honeyed pill lump combining process

Through near-infrared spectroscopy and moisture prediction model, the moisture detection problem in the process of Chinese medicine Miwan Hetuo was solved, and rapid and accurate moisture monitoring and control were achieved, and production efficiency and product stability were improved.

CN119023614APending Publication Date: 2024-11-26SHANDONG UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411076513.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-07
Publication Date
2024-11-26

AI Technical Summary

Technical Problem

The detection of moisture content during the process of Chinese medicine Mi Wan Hetuo is difficult to achieve rapid and accurate results, resulting in wasted time and energy during the production process. The conventional testing methods are complex and time-consuming, which affects production efficiency and product stability.

Method used

Using a moisture prediction model based on near-infrared spectrum, through near-infrared spectral scanning and moisture absorption treatment of the sample, a rapid detection model of moisture in the Chinese medicine Miwan Hetuo process was constructed, and combined with the model update strategy, real-time monitoring of the moisture content in the Chinese medicine Miwan Hetuo process was achieved. and control.

Benefits of technology

It realizes rapid and accurate detection, dynamic monitoring and control of the moisture content during the process of Chinese medicine Mi Wan Hetuo, reduces labor costs, improves production efficiency and product consistency, and reduces resource waste.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure BDA0004982428300000071
    Figure BDA0004982428300000071
  • Figure BDA0004982428300000081
    Figure BDA0004982428300000081
  • Figure BDA0004982428300000091
    Figure BDA0004982428300000091
Patent Text Reader

Abstract

The invention relates to a low-cost rapid detection method for moisture in a traditional Chinese medicine honeyed pill lump combining process, which comprises the following steps: drying a sample obtained by simulating the traditional Chinese medicine honeyed pill lump combining process, carrying out moisture absorption treatment, and carrying out spectrum detection on the traditional Chinese medicine honeyed pill lump combining sample after moisture absorption; correlating the obtained spectrum with the moisture content, constructing a model, then adopting a model updating method, selecting 10-20% of representative actual production process samples, adding the representative samples into a model training set, comparing model parameters, and screening an optimal processing mode, so as to dynamically monitor the moisture change by utilizing the updated model. According to the method, the moisture content change in the lump combining process can be rapidly detected in real time, so that the physical attributes in the lump combining process are reflected laterally, the sample collection and moisture content detection work in the conventional quantitative analysis model establishment process is simplified, the modeling time and workload are reduced, the labor cost is reduced, and the resource waste is reduced; and favorable development is provided for intelligent monitoring of the traditional Chinese medicine production process.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to a method for quickly detecting moisture in a low-cost Chinese medicine honey pill-making process, and belongs to the technical field of Chinese medicine moisture detection. Background Art

[0002] The process of combining Chinese medicine honey pills is a key step in the preparation of honey pills. The moisture content in the process of combining will affect the hardness and viscosity of the pills, which will directly affect the formability of the pills and the stability of the pills during storage. Therefore, by establishing a rapid detection model for moisture in the process of combining, it is possible to quickly correspond to the drying time of different standards, optimize the process production conditions, and obtain intermediates with appropriate hardness and softness for the next step of production or preservation. The process of combining Chinese medicine honey pills is a link with high energy consumption in the production of Chinese medicine. If the degree of combining is not enough, it will add unnecessary burden to the subsequent process, resulting in a waste of time and energy; if the degree of combining is too high, the viscosity of the intermediate pills is too high, which will adhere to the equipment too much, resulting in difficulty in transporting the pills. However, the amount of honey in the actual production process is certain, and sampling in the actual production process is time-consuming and labor-intensive, and the moisture content of the samples is generally consistent. The actual production cannot achieve process sampling with a large moisture content range. According to research, the moisture content of the lump-forming process is often used as an indicator to judge the physical properties of the lump, such as hardness, viscosity, plasticity, etc. Therefore, it is necessary to use the quantitative model of a large range of samples combined with the actual lump-forming samples of a small sample to predict the actual production process samples. However, conventional testing requires complex sampling steps and a long testing time, and directly using actual samples to build models requires a large sample size, which is likely to cause a waste of manpower, material and financial resources.

[0003] Therefore, a new method is urgently needed to reflect the changes in moisture content of the cakes in real time and accurately during the cake assembling process. Summary of the invention

[0004] In view of the deficiencies in the prior art, the present invention provides a low-cost method for quickly detecting moisture in the process of making Chinese medicine honey pills.

[0005] The low-cost method for rapid detection of moisture in the process of making honey pills of traditional Chinese medicine is based on a novel sample moisture prediction model, which realizes rapid determination of moisture content of real samples in the process of making Angong Niuhuang Pills, thereby making timely judgment on the endpoint of the process of making Angong Niuhuang Pills in terms of moisture content, realizing dynamic monitoring and precise control of moisture content of pills, and further achieving batch consistency.

[0006] The present invention is achieved through the following technical solutions:

[0007] A method for quickly detecting moisture in a Chinese medicine honey pill making process based on a moisture prediction model for the making process of Chinese medicine honey pills, the steps are as follows:

[0008] (1) performing near infrared spectroscopy scanning on samples of actual Chinese medicine honey pills and assembling process to obtain near infrared spectra of samples of actual Chinese medicine honey pills and assembling process;

[0009] (2) The moisture content of the actual samples of the Chinese medicine honey pills in the process of making honey pills was obtained by using the moisture prediction model of the Chinese medicine honey pills in the process of making honey pills.

[0010] The moisture prediction model of the Chinese medicine honey pill assembling process is constructed by the following method:

[0011] 1) Collect samples from the production process of Angong Niuhuang Pills to conduct drying weight loss, and use moisture absorption method to treat the original samples to obtain moisture absorption samples of Chinese medicine honey pills with different proportions of moisture content;

[0012] 2) Collect samples from the actual production process of Chinese medicine honey pills and measure the moisture content to obtain the actual moisture content of the actual production samples;

[0013] 3) Under the same testing conditions, the hygroscopic samples of the Chinese medicine honey pills and the samples of the Chinese medicine honey pills actually produced are subjected to near infrared spectroscopy scanning to obtain the near infrared spectra of the hygroscopic samples of the Chinese medicine honey pills and the samples of the Chinese medicine honey pills actually produced;

[0014] 4) effectively correlating the obtained near infrared spectra of the hygroscopic samples of the Chinese medicine honey pills with the moisture contents of the hygroscopic samples of the Chinese medicine honey pills with different moisture contents, and constructing a rapid determination model of moisture in the process of the Chinese medicine honey pills;

[0015] 5) The actual production of traditional Chinese medicine honey pills and combined samples were used as the external validation set, and the regression determination coefficient R of the external validation set was used. 2 t, the prediction root mean square error RMSET of the external validation set, and the performance ratio deviation RPDt of the external validation set are used to evaluate the prediction ability of the original model for the moisture content of samples in the actual process of making honey pills of traditional Chinese medicine;

[0016] 6) Adopting the active learning strategy of model updating, we compared three methods: random sampling RS, Kennard-Stone sampling KS, and X and Y spatial polynomial sampling SPXY. We selected 10-20% of the representative samples in the actual samples to add to the model training set, and used the remaining actual samples as the external validation set to examine the performance of the model after updating. 2 c. R 2 p, RMSEC, RMSEP, and RPD were used to evaluate the model stability performance. 2 The three parameters t, RMSET, and RPDt evaluate the stability and extrapolation ability of the updated model;

[0017] 7) The near-infrared spectra of the obtained hygroscopic samples of Chinese medicine honey pills and the near-infrared spectra of representative samples of the actual production of Chinese medicine honey pills are effectively correlated with the moisture content of the hygroscopic samples of Chinese medicine honey pills with different moisture contents and the corresponding moisture content of the actual samples, and a moisture prediction model for the Chinese medicine honey pill process is constructed.

[0018] Preferably according to the present invention, in step 1), the moisture absorption method is to perform moisture absorption treatment on the dry sample using a saturated ammonium persulfate solution in a closed container.

[0019] Preferably, according to the present invention, in step 1), the moisture content of the hygroscopic sample of the Chinese medicine honey pills is determined to be in the range of 1-20wt%.

[0020] Preferably according to the present invention, in step 4), the effective association is to assign the moisture content of the sample to the corresponding near-infrared spectrum according to the characteristic absorption peak of the hydroxyl group (OH) in the water molecule in the near-infrared region, the near-infrared spectrum of each sample corresponds to a moisture value, and then use a chemometric method to establish a quantitative analysis model based on the characteristic wavelength and the known moisture content.

[0021] Preferably, according to the present invention, in step 4), the model obtained by associating the near-infrared spectrum with the moisture content adopts the linear correction method of partial least squares regression PLSR, and when selecting representative samples of the actual Chinese medicine honey pill making process, RS, KS, and SPXY methods are used for comparison, and finally the KS method is determined to obtain the best updated model, so as to further make the model suitable for new data distribution and application scenarios.

[0022] Preferably, in step 6), the active learning strategy of model updating is as follows:

[0023] a. Determine the update goal: clarify the purpose of model update, including improving the prediction accuracy of the model, adapting to new sample distribution, and enhancing the generalization ability of the model;

[0024] b. Collect new data: Collect new sample data according to the updated target; the new sample data comes from new experiments, field measurements or different data sources;

[0025] c. Preliminary analysis: Conduct preliminary analysis on the newly collected data. Preliminary analysis includes data screening and exploratory data analysis to understand the characteristics and potential problems of the new data;

[0026] d. Active learning strategy: Design active learning strategies to select the most informative samples for model updating, including uncertainty sampling, boundary exploration, and diversity sampling;

[0027] e. Model evaluation: Use the existing model to predict new data and evaluate the model performance to determine the shortcomings of the model on the new data;

[0028] f. Model update: Based on the evaluation results, select an appropriate model update method, which includes incremental learning: gradually integrating new data into the existing model and updating model parameters;

[0029] g. Model retraining: Retrain the model using the selected samples and strategies;

[0030] h. Verification and testing: Use an independent test set to verify the performance of the updated model; evaluate whether the model update has achieved the expected goals.

[0031] i. Iterative optimization: Based on the verification results, the model update process needs to be further iterated and optimized, including adjusting the active learning strategy, selecting different samples, or trying different model update methods;

[0032] j. Knowledge integration: Integrate newly learned knowledge into the existing knowledge base to ensure that the model can adapt to new data distribution and application scenarios.

[0033] Preferably, according to the present invention, in step d, uncertainty sampling: select the samples with the most uncertainty in model prediction for labeling and learning; boundary exploration: select samples located near the decision boundary; diversity sampling: ensure that the selected samples are representative and diverse in the feature space; in step d, samples that can represent the distribution of new scene data can also be selected by comparison through KS, RS, and SPXY methods.

[0034] According to the preferred embodiment of the present invention, in step e, R 2 t, RPDt, and RMSET parameters are used to evaluate model performance.

[0035] The present invention adopts the model updating active learning strategy to effectively update the model, make it better adapt to new data and tasks, and improve the practicality and accuracy of the model.

[0036] Preferably, according to the present invention, in step 6), the model training set is a data set for constructing and training a machine learning or statistical model.

[0037] Preferably, according to the present invention, in step 6), when the performance of the original model for the new scenario is poor and the model cannot make accurate predictions, representative samples in the new scenario are added to the model training set, following the principle of adding the least new samples and improving model performance the best, to improve the prediction performance of the original model for the new scenario samples.

[0038] Near infrared spectroscopy cannot be directly used for the detection of moisture content in the process of making honey pills of traditional Chinese medicine. The present invention uses real-time, rapid, and green near infrared spectroscopy to effectively associate the spectrum obtained by the detection with the moisture content using chemometric means, and establish a quantitative detection model by association; thereby realizing the pairing of the spectrum of the unknown sample with the model, so as to obtain the moisture content of the sample in the process of making honey pills of traditional Chinese medicine corresponding to the spectrum, and realizing the rapid detection of moisture in the process of making honey pills of traditional Chinese medicine. In addition, the spectrum detection device can be installed on the mixing machine in the form of a detector, and interconnected with the model stored in the computer to detect the moisture content in the process of making honey pills of traditional Chinese medicine in real time, so as to realize the dynamic monitoring of the moisture content of the pills on the one hand, and judge the moisture content at the end point of making pills on the other hand, accurately grasp the consistency between batches, and obtain the maximum benefit with the minimum labor cost.

[0039] Since samples of the process of making Chinese medicine honey pills cannot be obtained at will and are difficult to obtain, the moisture content in the process of making Chinese medicine honey pills is difficult to obtain. The present invention uses a simulated production process of making Chinese medicine honey pills to collect samples in the process, and further dry and absorb moisture to obtain samples of Chinese medicine honey pills with different moisture contents, ensuring that the moisture-absorbed samples cover the moisture content of the actual production samples. However, due to differences in usage scenarios, the model established by laboratory simulation samples has poor robustness when predicting the moisture content of actual production samples. Therefore, the present invention introduces representative samples in actual production to participate in the model training set, thereby improving the extrapolation ability and robustness of the model, thereby eliminating differences in scenario applicability.

[0040] The technical features and advantages of the present invention are as follows:

[0041] 1. The low-cost method for rapid detection of moisture in the process of assembling honey pills of traditional Chinese medicine of the present invention utilizes a model for effective association of the constructed spectrum, representative actual sample spectrum and moisture content, and realizes rapid acquisition of moisture content by spectral detection of samples of the process of assembling honey pills of traditional Chinese medicine to be tested, thereby realizing dynamic monitoring of moisture in the assembling process and determining moisture content at the assembling endpoint, which is beneficial to the next process and realizes prediction of actual production process samples by using a quantitative model of a large range of samples combined with actual assembling samples of a small sample.

[0042] 2. The low-cost method for rapid detection of moisture in the process of making honey pills of traditional Chinese medicine adopts drying and moisture absorption methods to obtain samples of the process of making honey pills of traditional Chinese medicine with different moisture ratios, so that the sample set covers the sample conditions that may occur in the actual process of making pills. The method avoids the difficulty of sampling in conventional actual production, solves the problems of sample collection and overtime detection of moisture content in the process of establishing conventional quantitative analysis models, reduces the time and workload required for modeling, reduces labor costs, and reduces waste of resources;

[0043] 3. The present invention introduces actual production process samples through model updating to improve the scope of application of the model, thereby improving the model's predictive ability for actual samples and increasing the feasibility of online application of the model, thereby ensuring the accuracy and real-time nature of moisture content obtained by detecting near-infrared spectroscopy, providing favorable development for intelligent monitoring of Chinese medicine production processes. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 It is the original near infrared spectra of the Angong Niuhuang Pills samples and the actual production process samples after moisture absorption for different times in the embodiment of the present invention, a is the original near infrared spectra of the Angong Niuhuang Pills with different moisture contents, and b is the original near infrared spectra of the actual production process samples;

[0045] Figure 2 It is a PCA score distribution diagram of Angong Niuhuang Pill samples after moisture absorption at different times and samples in the actual production process in the embodiment of the present invention;

[0046] Figure 3 The moisture quantitative analysis model of the samples of the Angong Niuhuang Pills Hetuan process after different moisture absorption times in the embodiment of the present invention and the comparison diagram of the model prediction and the reference value; a is the analysis model, and b is the model prediction and the reference value;

[0047] Figure 4 This is a comparison chart between the model prediction and reference value of the moisture content of samples in the actual production process of lump-forming by the original moisture quantitative model in the embodiment of the present invention;

[0048] Figure 5 : This is a score distribution diagram of the representative samples selected by KS and SPXY in the embodiment of the present invention in the actual samples, where a is a representative sample selected by KS represented by △, b is two representative samples selected by SPXY represented by *, and c is five representative samples selected by SPXY represented by ◇;

[0049] Figure 6 1 is a comparison diagram of the reference value and the model prediction value before and after the RS, KS and SPXY partition and update the model in the embodiment of the present invention, a is a comparison diagram of the reference value and the average model prediction value after RS ​​randomly selects 20 groups of samples, b is a comparison diagram of the reference value and the model prediction value after KS selects the best number of samples, c is a comparison diagram of the reference value and the model prediction value after SPXY selects 10% of the samples, and d is a comparison diagram of the reference value and the model prediction value after SPXY selects 20% of the samples;

[0050] Figure 7 1 is a model for quantitative analysis of moisture in the process of combining Angong Niuhuang Pills and its external validation diagram after the model is updated in an embodiment of the present invention. a is a model diagram for quantitative analysis of moisture in the process of combining Angong Niuhuang Pills after the model is updated, and b is an external validation diagram of the updated model. DETAILED DESCRIPTION

[0051] It should be noted that the following detailed description is exemplary and is intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those of ordinary skill in the art to which the present invention belongs. The present invention is further explained below by specific examples, but the protection scope of the present invention is not limited thereto.

[0052] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0053] In order to enable those skilled in the art to clearly understand the technical solution of the present invention, the technical solution of the present invention will be described in detail below in conjunction with an embodiment of a specific Angong Niuhuang Pills compounding process.

[0054] Example 1

[0055] A method for quickly detecting moisture in a Chinese medicine honey pill making process based on a moisture prediction model for the making process of Chinese medicine honey pills, the steps are as follows:

[0056] (1) Sample preparation

[0057] Samples of the simulated Angong Niuhuang Pill production process were collected and placed in a vacuum drying oven at 105°C to obtain dried (moisture content ≈ 0) samples. At the same time, the dried samples were subjected to moisture absorption using a saturated ammonium persulfate solution in a closed container, and it was determined that the moisture absorption range of the samples was approximately 1-20wt%, and the moisture absorption rate of the samples during the moisture absorption process was first fast and then slow, and finally maintained a stable state; finally, by controlling the time of the dried samples in the closed container, moisture-absorbed samples with different moisture content ratios were obtained, and the original near-infrared spectra of the Angong Niuhuang Pill samples after moisture absorption for different times and the samples in the actual production process were shown in Figure 2. Figure 1 , expand the moisture content range to ensure that its moisture content range covers the moisture content of combined samples that may appear in actual production. Each batch of experiments takes 21 time points and is repeated three times. There are two groups of experiments with 4 samples each, and a total of 65 experimental samples;

[0058] At the same time, samples from the actual production process of Angong Niuhuang Pills were collected as external verification, with a total of 27 samples.

[0059] (2) Spectral acquisition

[0060] Antaris II Fourier transform near infrared spectrometer was used to collect near infrared spectra of hygroscopic lump samples and actual production lump samples. Air was used as background reference and background collection was performed every 5 minutes. The samples were placed in a suitable collection accessory and placed in a fixed sample pool. Spectral collection was performed in a diffuse reflectance manner. The spectral scanning parameters were set to: scanning range 4000-10000cm -1 , the number of scans is 64 times, the resolution is 8cm -1 After each sample is collected three times, the system automatically calculates its average value and uses its average spectrum as the original spectrum of each sample. The original spectrum scanning results of the hygroscopic combined lump sample and the actual production combined lump sample are shown in Figure 1 , and its PCA score distribution is shown in Figure 2 ;

[0061] (3) Determination of moisture content

[0062] a Determination of moisture content in hygroscopic lump samples

[0063] The moisture content of the sample is characterized by the weight gain method. The specific formula is shown in Formula 1. The moisture content of the laboratory combined sample determined by the weight gain method is shown in Table 1;

[0064] Moisture content % = (m3-m2) / [(m3-m2)+m1] Formula 1;

[0065] Among them, m1 is the mass of the dry sample, m2 is the total mass of the glass dish and the dry sample, and m3 is the total mass of the glass dish and the dry sample after moisture absorption. The formula shows how much the weight of the sample increases after moisture absorption accounts for the original weight of the sample; the moisture content of the combined sample at different moisture absorption times in the laboratory is obtained by calculation.

[0066] Table 1 Moisture content of Angong Niuhuang Pills moisture absorption combined sample

[0067]

[0068]

[0069]

[0070] Samples 1, 2, and 3 in the table are 3 parallel samples at each time.

[0071] b. Determination of moisture content of combined samples in actual production process

[0072] Combined with the actual conditions of the factory, the moisture content of the samples was directly measured using a moisture meter (DSH-10A) after the collection was completed, and the moisture content data was shown in Table 2;

[0073] Table 2 Moisture content of combined samples in the actual production process of Angong Niuhuang Pills

[0074]

[0075]

[0076] (4) Establishment of quantitative analysis model

[0077] The PLSR method was used to establish a quantitative analysis model for the moisture content in the combined samples at different moisture absorption times.

[0078] Specific methods: ① The laboratory moisture absorption samples were divided into model calibration and validation sets in a ratio of 2:1 using the KS partitioning method; ② The calibration set determination coefficient (R 2 c) Cross-validation set determination coefficient (R 2 cv), validation set determination coefficient (R 2 v), corrected root mean square error (RMSEC), cross validation set root mean square error (RMSECV), validation set root mean square error (RMSEV) and performance deviation ratio (RPD) to evaluate the prediction ability of the model; ③ Based on the original model, nine methods including MC, Auto, MSC, SNV, Normalize smoothing, SG smoothing, derivative and CWT were used alone or in combination to optimize the original spectrum. At the same time, RMSEV, R 2 v and RPD are the evaluation indicators. The larger the RPD value, the smaller the RMSEP value, and the 2 The principle of a larger p value is to comprehensively screen the optimal spectral preprocessing method to eliminate the influence of uneven distribution of solid particles and surface scattering on diffuse reflectance spectra; ④ Select the optimal band. The spectral interval selected for modeling should contain the maximum amount of water information and potential correlation and interaction effects, and reduce noise interference as much as possible, so as to make full use of the information of the entire spectral range, and thus provide better prediction performance and generalization ability. The results of the moisture quantitative analysis model are shown in Figure 3 ,It can be seen from the figure that the predicted value and the reference value have a good fit. The evaluation parameters of the moisture quantitative analysis model are shown in Table 3. The R 2 They are all greater than 0.97, and the RPD value is greater than 4, which indicates that the model has good predictive ability and can accurately predict the moisture content of Hetuo samples.

[0079] Table 3 Results of evaluation parameters of the quantitative analysis model for moisture in Angong Niuhuang Wan Hetuo samples

[0080]

[0081] Note: Q 2 The variance explanation rate of the principal components is generally required to be greater than 0.8.

[0082] (5) Application of quantitative analysis model

[0083] The samples of the actual production process were used as the external validation set, and the original model was used to quickly predict the moisture content of the samples of the actual production process. The prediction results are as follows: Figure 4 As shown in the figure, it can be seen that the model prediction results deviate greatly from the actual measurement results, and the original model cannot accurately predict the actual production process samples. At the same time, a t-test was performed, and the p value of 2.7011e-13 was less than the threshold of 0.01, indicating that there was a very significant difference between the two groups of data.

[0084] (6) Selecting a representative sample

[0085] The results of external validation of the model by adding the selected representative samples were evaluated using random sampling (RS), KS (Kennard stone, KS) sampling, and SPXY sampling methods. The most representative samples of 10-20% of the external validation set were selected and added to the calibration set to update the quantitative model with RMSET, R 2 t and RPDt are evaluation indicators, and the fewer the number of external validation sets added, the closer RMSET is to 0, and R 2 The closer t is to 1 and the larger the RPDt value is, the more representative samples are selected, and the performance of the updated model is evaluated based on this, so that the model has better extrapolation and robust performance when predicting the moisture content of actual samples. The results of the model evaluation parameters of the three selection methods are shown in Table 4; at the same time, the score distribution of the representative samples selected by KS and SPXY in the actual samples is shown in the figure Figure 5 As shown in the figure, the comparison between the reference value and the predicted value after the three partitioning methods update the model is as follows: Figure 6 shown.

[0086] Table 4 Evaluation parameter results of the three selection methods for updating the moisture quantitative analysis model

[0087]

[0088]

[0089] (7) Model update application

[0090] The updated model was applied to the sample moisture prediction in the actual production process. The updated model and its prediction results of the moisture quantitative analysis model of the Angong Niuhuang Pills process are shown in Figure 2. Figure 7 As shown in the figure, it can be seen that the updated model has a good fitting effect, and the prediction results are basically consistent with the actual measurement results, which can meet the dynamic monitoring of the moisture content of the samples during the merging process.

[0091] The above results show that the rapid detection method of moisture in the process of merging based on the moisture prediction model of the process of merging of Chinese medicine honey pills of the present invention can quickly detect the moisture content of samples in the process of merging of Chinese medicine honey pills, thereby dynamically monitoring the moisture changes and providing a basis for judging the moisture content of the end point of merging. The invention uses a laboratory simulation of the process of merging of Chinese medicine honey pills to collect samples in the process, further dry and absorb moisture, and obtain samples of Chinese medicine honey pills with different moisture contents, so as to ensure that the laboratory moisture absorption samples cover the moisture content of the actual production samples, and use the model update angle to add representative actual production samples to the model training set, thereby simplifying the sample collection and moisture content detection work in the process of establishing the conventional quantitative analysis model, and also improving the hysteresis problem of moisture content detection, reducing the time and workload required for modeling, reducing labor costs, and reducing resource waste; at the same time, the actual production process samples are introduced to improve the scope of application of the model, thereby improving the model's prediction ability for new scene samples, and increasing the feasibility of the model for online application, thereby ensuring the accuracy and real-time of moisture content obtained by detecting near-infrared spectroscopy, and providing favorable development for intelligent monitoring of Chinese medicine production process.

[0092] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for rapid detection of moisture in the process of making a honey pill of traditional Chinese medicine based on a moisture prediction model of the process of making a honey pill of traditional Chinese medicine, the steps are as follows: (1) performing near infrared spectroscopy scanning on samples of actual Chinese medicine honey pills and assembling process to obtain near infrared spectra of samples of actual Chinese medicine honey pills and assembling process; (2) The moisture content of the samples of the actual Chinese medicine honey pills in the process of making the pills is obtained by using the moisture prediction model of the Chinese medicine honey pills in the process of making the pills; The moisture prediction model of the Chinese medicine honey pill assembling process is constructed by the following method: 1) Collect samples from the production process of Angong Niuhuang Pills to conduct drying weight loss, and use moisture absorption method to treat the original samples to obtain moisture absorption samples of Chinese medicine honey pills with different proportions of moisture content; 2) Collect samples from the actual production process of Chinese medicine honey pills and measure the moisture content to obtain the actual moisture content of the actual production samples; 3) Under the same testing conditions, the hygroscopic samples of the Chinese medicine honey pills and the samples of the Chinese medicine honey pills actually produced are subjected to near infrared spectroscopy scanning to obtain the near infrared spectra of the hygroscopic samples of the Chinese medicine honey pills and the samples of the Chinese medicine honey pills actually produced; 4) effectively correlating the obtained near infrared spectra of the hygroscopic samples of the Chinese medicine honey pills with the moisture contents of the hygroscopic samples of the Chinese medicine honey pills with different moisture contents, and constructing a rapid determination model of moisture in the process of the Chinese medicine honey pills; 5) The actual production of traditional Chinese medicine honey pills and combined samples were used as the external validation set, and the regression determination coefficient R of the external validation set was used. 2 t, the prediction root mean square error RMSET of the external validation set, and the performance ratio deviation RPDt of the external validation set are used to evaluate the prediction ability of the original model for the moisture content of samples in the actual process of making honey pills of traditional Chinese medicine; 6) Adopting the active learning strategy of model updating, we compared three methods: random sampling RS, Kennard-Stone sampling KS, and X and Y space polynomial sampling SPXY. We selected 10-20% of the representative samples in the actual samples to add to the model training set, and used the remaining actual samples as the external validation set to examine the performance of the model after updating. We used R2c, R2p, RMSEC, RMSEP, and RPD to evaluate the model stability performance, and used R2t, RMSET, and RPDt to evaluate the stability and extrapolation ability of the updated model. 7) The near-infrared spectra of the obtained hygroscopic samples of Chinese medicine honey pills and the near-infrared spectra of representative samples of the actual production of Chinese medicine honey pills are effectively correlated with the moisture content of the hygroscopic samples of Chinese medicine honey pills with different moisture contents and the corresponding moisture content of the actual samples, and a moisture prediction model for the Chinese medicine honey pill process is constructed.

2. The rapid detection method according to claim 1, It is characterized in that In step 1), the moisture absorption method is to use a saturated ammonium persulfate solution in a closed container to absorb moisture on the dry sample, and determine that the moisture content of the Chinese medicine honey pill moisture absorption sample is in the range of 1-20wt%.

3. The rapid detection method according to claim 1, It is characterized in that In step 4), the effective association is to assign the moisture content of the sample to the corresponding near-infrared spectrum according to the characteristic absorption peak of the hydroxyl group (OH) in the water molecule in the near-infrared region. The near-infrared spectrum of each sample corresponds to a moisture value, and then a chemometric method is used to establish a quantitative analysis model based on the characteristic wavelength and the known moisture content.

4. The rapid detection method according to claim 1, It is characterized in that In step 4), the model obtained by associating the near-infrared spectrum with the moisture content adopts the linear correction method of partial least squares regression (PLSR). When selecting representative samples of the actual Chinese medicine honey pill making process, the RS, KS, and SPXY methods are used for comparison. Finally, the KS method is determined to obtain the best updated model, which further makes the model suitable for new data distributions and application scenarios.

5. The rapid detection method according to claim 1, It is characterized in that In step 6), the active learning strategy using model update is as follows: a. Determine the update goal: clarify the purpose of model update, including improving the prediction accuracy of the model, adapting to new sample distribution, and enhancing the generalization ability of the model; b. Collect new data: Collect new sample data according to the updated target; the new sample data comes from new experiments, field measurements or different data sources; c. Preliminary analysis: Conduct preliminary analysis on the newly collected data. Preliminary analysis includes data screening and exploratory data analysis to understand the characteristics and potential problems of the new data; d. Active learning strategy: Design active learning strategies to select the most informative samples for model updating, including uncertainty sampling, boundary exploration, and diversity sampling; e. Model evaluation: Use the existing model to predict new data and evaluate the model performance to determine the shortcomings of the model on the new data; f. Model update: Based on the evaluation results, select an appropriate model update method, which includes incremental learning: gradually integrating new data into the existing model and updating model parameters; g. Model retraining: Retrain the model using the selected samples and strategies; h. Verification and testing: Use an independent test set to verify the performance of the updated model; evaluate whether the model update has achieved the expected goals; i. Iterative optimization: Based on the verification results, the model update process needs to be further iterated and optimized, including adjusting the active learning strategy, selecting different samples, or trying different model update methods; j. Knowledge integration: Integrate newly learned knowledge into the existing knowledge base to ensure that the model can adapt to new data distribution and application scenarios.

6. The rapid detection method according to claim 5, It is characterized in that In step d, uncertainty sampling: select the samples with the most uncertain model predictions for labeling and learning; boundary exploration: select samples near the decision boundary; diversity sampling: ensure that the selected samples are representative and diverse in the feature space.

7. The rapid detection method according to claim 5, It is characterized in that In step d, samples that can represent the data distribution of the new scene can also be selected by comparison using the KS, RS, and SPXY methods.

8. The rapid detection method according to claim 5, It is characterized in that In step e, R 2 t, RPDt, and RMSET parameters are used to evaluate model performance.

9. The rapid detection method according to claim 1, It is characterized in that In step 6), the model training set is a data set for building and training a machine learning or statistical model.

10. The rapid detection method according to claim 1, It is characterized in that In step 6), when the original model performs poorly for the new scenario and cannot make accurate predictions, representative samples in the new scenario are added to the model training set, following the principle of adding the least new samples and improving model performance to improve the prediction performance of the original model for the new scenario samples.