Model prediction optimization-based dogwood health wine soaking process

Abstract

The invention provides a dogwood health wine soaking process based on model prediction optimization, and relates to the technical field of dogwood health wine soaking processes, and the dogwood health wine soaking process comprises the following steps: calculating a comprehensive cell wall degradation index, grading raw materials, and coding the graded raw materials into raw material quality grade values; establishing a first-order dissolution kinetic model, and performing parameter optimization in combination with a golden section search algorithm to obtain an optimal infusion temperature and an optimal infusion time; weighing the medicinal materials according to a preset ratio, mixing the medicinal materials with the base liquor, and starting to soak; arranging a two-parameter collaborative monitoring system in the soaking container; constructing a random forest model, predicting the dissolution completion degree and the predicted residual equilibrium time, performing dynamic regulation and control, and ending the dynamic regulation and control process when the prediction result meets the requirement to obtain an immersion liquid; and taking out the soaking liquid from the soaking container, and separating and purifying by adopting a secondary filtering process to obtain the dogwood health wine product. The technical defects that a traditional process depends on experience, parameters are solidified, and the monitoring means is single can be overcome.

Description

Technical Field

[0001] This invention relates to the field of Cornus officinalis health wine infusion technology, and in particular to a Cornus officinalis health wine infusion process based on model prediction optimization. Background Technology

[0002] Cornus officinalis health-preserving wine, as an important category of traditional medicinal wine, is mainly prepared using the baijiu (Chinese white liquor) extraction method. Current baijiu health-preserving wine immersion processes typically include raw material pretreatment, mixing and immersion of medicinal materials with base liquor, and filtration and clarification. Regarding immersion parameters, traditional processes often rely on empirically established fixed parameters. During the immersion process of Cornus officinalis and other Chinese medicinal materials, the dissolution of active ingredients such as iridoid glycosides and organic acids mainly relies on diffusion mass transfer. Process parameters such as immersion temperature, time, and alcohol concentration directly affect the dissolution efficiency of effective components and product quality.

[0003] The main technical problems in the current process of extracting Cornus officinalis health wine include: a lack of scientific basis for setting process parameters, which are mostly determined by experience, leading to large fluctuations in the quality of different batches of products; a lack of effective methods for assessing differences in raw material quality, making it difficult to adjust process parameters according to the properties of the raw materials; and limited monitoring methods for the extraction process, mainly relying on time control, which makes it impossible to judge the completion of dissolution in real time. To address these problems, some studies have attempted to optimize the extraction process parameters using statistical methods such as orthogonal experiments and response surface methodology, but these remain at the static optimization stage and have not yet resulted in a dynamic control solution.

[0004] Chinese invention patent CN106929363A discloses a health-preserving wine and its preparation method. This patent employs a graded soaking process, placing Epimedium at the bottom of a soaking tank with other medicinal materials on top. It first soaks the herbs in 65% baijiu (Chinese white liquor) for 15-25 days, then in 45% baijiu for 10-20 days, and finally in purified water. The mixture is circulated for 4-6 hours daily during the soaking process, and the herb ratio is optimized using response surface methodology. However, this patent does not address the impact of raw material quality differences on the selection of process parameters and lacks dynamic parameter adjustment technology based on raw material properties. Summary of the Invention

[0005] In view of this, the present invention provides a model prediction optimization-based process for the maceration of Cornus officinalis health wine. By establishing a cell wall degradation index evaluation system to quantitatively evaluate the properties of raw materials, a first-order dissolution kinetic model combined with the golden section search algorithm is used to scientifically determine process parameters. A random forest model is used to construct a dual-parameter collaborative monitoring system to achieve intelligent dynamic control of the maceration process. This solves the technical defects of traditional processes that rely on experience, have fixed parameters, and have limited monitoring methods, and achieves intelligent optimization of the Cornus officinalis health wine production process and stable control of product quality.

[0006] The technical solution of this invention is implemented as follows:

[0007] This invention provides a model-predictive optimized process for making Cornus officinalis health-preserving wine, comprising:

[0008] S1. The active ingredients of five medicinal materials, namely Cornus officinalis, mulberry, Poria cocos, jujube and ginseng, are tested using the corresponding standard methods. A cell wall degradation index evaluation system is established through mechanical strength testing. The cell wall degradation index of each medicinal material is measured and the comprehensive cell wall degradation index is calculated by preset ratio weights. A comprehensive quality evaluation function is established based on the content of active ingredients and the comprehensive cell wall degradation index. The raw materials are graded and coded as raw material quality grade values.

[0009] S2. Based on the raw material quality grade value and the comprehensive cell wall degradation index, a first-order dissolution kinetic model based on the equivalent active ingredient concentration is established. The modified Arrhenius equation is used to describe the relationship between the dissolution rate constant and the temperature and raw material properties. Based on this dissolution kinetic model, the single-objective parameter optimization is carried out in combination with the golden section search algorithm to obtain the corresponding optimal immersion temperature and optimal immersion time.

[0010] S3. Weigh the medicinal materials and mix them with the base liquor according to the preset ratio. After drying and pulverizing the medicinal materials, put them into the soaking container and start soaking according to the optimal soaking temperature and optimal soaking time.

[0011] S4. Install pH and temperature sensors in the immersion container to establish a dual-parameter collaborative monitoring system, and adopt a phased monitoring strategy to collect monitoring parameters in real time during the immersion process.

[0012] S5. Construct a random forest model, using the rate of change of monitoring parameters, immersion time, raw material quality grade, comprehensive cell wall degradation index, and pH-temperature synergistic index as input features. Simultaneously predict the dissolution completion rate and the estimated remaining equilibrium time. Based on the prediction results, dynamically adjust the temperature and time parameters in the immersion container. When the predicted dissolution completion rate reaches 95% and the estimated remaining equilibrium time is ≤1 day, end the dynamic adjustment process, and continue to wait for the predicted remaining equilibrium time to complete the immersion to obtain the immersion solution.

[0013] S6. Remove the maceration solution from the maceration container, separate and purify it using a two-stage filtration process, and conduct quality testing and verification on the product to obtain the Cornus officinalis health wine product.

[0014] The preferred formula for calculating the comprehensive quality evaluation function is as follows:

[0015]

[0016] Among them, Q total The overall quality score of the medicinal materials; C iThe content of the i-th active ingredient includes the content of mononosides and organic acids in Cornus officinalis, the content of anthocyanins and total sugars in Morus alba, the content of polysaccharides in Poria cocos, the content of total sugars in Jujube, and the content of saponins in Ginseng; α i The difficulty index for dissolving active ingredients; W i R represents the importance weighting coefficient. barrier,i β is the cell wall diffusion resistance constant of the i-th active ingredient; β is the cell wall deterioration and dissolution promotion coefficient. This is a comprehensive cell wall degradation index.

[0017] Preferably, the cell wall degradation index is quantified through standardized mechanical strength testing. The formula for calculating the cell wall degradation index of a single medicinal material is as follows:

[0018]

[0019] Where j = 1, 2, 3, 4, 5 correspond to Cornus officinalis, mulberry, Poria cocos, jujube, and ginseng, respectively; F base,j F is the benchmark crushing force for the corresponding medicinal materials. measured,j For actual crushing force;

[0020] The overall cell wall degradation index is calculated using a weighted average:

[0021]

[0022] Where w j The weighting coefficients for each medicinal material are determined based on the preset proportions.

[0023] The preferred calculation formula for the first-order dissolution kinetic model is as follows:

[0024] C(t)=C max ·[1-exp(-k·t)]

[0025] Where C(t) is the equivalent active ingredient concentration at time t; C max The theoretical maximum equivalent dissolution concentration is given by k; the overall dissolution rate constant is given by t; and the immersion time is given by t.

[0026] The relationship between the overall dissolution rate constant k and temperature and overall feed properties follows the modified Arrhenius equation:

[0027]

[0028] Where k0 is the reference rate constant, E a γ is the apparent activation energy, R is the gas constant, T is the absolute temperature, γ is the cell wall degradation promoting coefficient, μ is the quality grade correction coefficient, and G is the apparent activation energy. raw This represents the raw material quality grade value.

[0029] Preferably, the specific optimization process of the golden section search algorithm includes:

[0030] Step 1: Set search parameters. According to the raw material quality grade value G raw and the comprehensive cell wall deterioration index set the temperature search range [T min , T max and the time search range [t min , t max , set the precision requirement ∈ T and ∈ t ;

[0031] Step 2: Construct an optimization objective function:

[0032]

[0033] The goal is to find the optimal temperature-time combination that makes the dissolution completion rate reach 95%;

[0034] Step 3: Calculate the initial objective function value. Select two golden section points T1 = T min + 0.382×(T max - T min ) and T2 = T min + 0.618×(T max - T min ) within the temperature search interval, and calculate the objective function values F(T1, t) and F(T2, t) respectively;

[0035] Step 4: Perform optimization iteration in the temperature dimension. When F(T1, t) < F(T2, t), update the search interval to [T min , T2] and reselect the golden section point. When F(T1, t) ≥ F(T2, t), update the search interval to [T1, T max and reselect the golden section point;

[0036] Step 5: Perform optimization iteration in the time dimension. Fix the optimal temperature obtained in Step 4 and use the same golden section method to find the optimal time parameter within the time search interval;

[0037] Step 6: Alternately execute Step 4 and Step 5. When the changes in the temperature and time parameters for two consecutive iterations are both less than the precision requirement or the maximum number of iterations is reached, end the optimization process and output the optimal dipping temperature T opt and the optimal dipping time t opt .

[0038] Preferably, the method further includes: establishing a parameter quick query table, and matching m1 levels of the raw material quality grade G raw with the comprehensive cell wall deterioration index The pre-defined m2 intervals are combined to form m1×m2 combinations of raw material properties; the optimal immersion temperature T for each combination is pre-calculated using the golden section search algorithm. opt and optimal immersion time t opt Establish and store a parameter lookup table; during production, directly query the corresponding process parameters based on the raw material quality grade value and comprehensive cell wall degradation index from step S1; for raw material property parameters that are between the middle values ​​of the preset range, use bilinear interpolation to calculate the corresponding temperature and time parameters.

[0039] Preferably, the random forest model includes a first sub-model and a second sub-model; during prediction, the two sub-models are called simultaneously to process the same input feature vector, and the prediction results of dissolution completion and remaining equilibrium time are obtained respectively. The prediction results of the two sub-models are fused by dynamic weights based on the current immersion time to obtain the final prediction result.

[0040] Preferably, in step S5, the dynamic control strategy is as follows:

[0041] When the dissolution process is delayed, the immersion time is extended; when the pH fluctuates abnormally, the immersion temperature is adjusted to stabilize the chemical equilibrium; when the predicted remaining time is abnormal, the temperature and time are adjusted in a coordinated manner; when the predicted dissolution completion rate reaches 95% and the expected remaining equilibrium time is ≤1 day, the dynamic control process ends and the immersion continues to wait for the predicted remaining equilibrium time to complete.

[0042] Preferably, the preset ratio in step S3 is the weight ratio m of Cornus officinalis, mulberry, Poria cocos, jujube, and ginseng. 山茱萸 :m 桑椹 :m 茯苓 :m 大枣 :m 人参 = 5:5:4:5:1; the base liquor is 53-degree soy sauce-flavored baijiu, which is mixed with the medicinal materials at a mass ratio of 1:10; the pretreatment of the medicinal materials includes drying at a constant temperature of 60℃ until the moisture content is ≤10% and crushing through a 20-40 mesh sieve.

[0043] Preferably, in step S6, the secondary filtration process includes: primary coarse filtration using a 200-mesh stainless steel filter to remove medicinal material particles and large molecular impurities, with the filtration rate controlled at 50-80 mL / min; and secondary fine filtration using a 0.45 μm ceramic membrane filter to ensure product clarity.

[0044] Quality testing and verification include color a* value testing, clarity NTU value testing, dry matter content testing, key active ingredient content testing, and sensory evaluation; a quality consistency evaluation system is established, and process stability is assessed by calculating the batch-to-batch coefficient of variation for each test indicator, with a target coefficient of variation control of less than 5%.

[0045] The present invention has the following advantages over the prior art:

[0046] (1) This invention establishes a cell wall degradation index evaluation system and a first-order dissolution kinetic model based on the concentration of equivalent active ingredients, combines the golden section search algorithm to optimize process parameters, and uses a random forest model to construct a dual-parameter collaborative monitoring system to realize intelligent dynamic control of the immersion process. This solves the technical problems in the traditional Cornus officinalis health wine immersion process, such as the lack of scientific basis for setting process parameters, large fluctuations in the quality of different batches of products, lack of effective assessment of raw material quality differences, and limited means of monitoring the immersion process.

[0047] (2) The cell wall degradation index evaluation system established in this invention quantitatively evaluates the degree of cell wall degradation of medicinal materials through standardized mechanical strength testing, providing an objective basis for the scientific assessment of raw material quality differences. By measuring the crushing force of each medicinal material under standard conditions and comparing it with the benchmark value, the integrity of the cell wall structure can be accurately reflected, thereby predicting the dissolution difficulty of active ingredients, and transforming the setting of process parameters from empirical judgment to quantitative analysis based on physical properties;

[0048] (3) The first-order dissolution kinetic model based on equivalent active ingredient concentration combined with the modified Arrhenius equation adopted in this invention can accurately describe the comprehensive influence of temperature, raw material quality and cell wall degradation on the dissolution rate. By using the golden section search algorithm for single-objective parameter optimization, the optimal combination of temperature and time that achieves 95% dissolution completion can be quickly determined, avoiding the complexity and parameter conflict problems of traditional multi-objective optimization;

[0049] (4) The random forest model constructed in this invention adopts a phased dual-task learning framework. By using an early sub-model to handle the rapid dissolution period characteristics and a mid-to-late-stage sub-model to handle the equilibrium dissolution period characteristics, it can simultaneously predict the dissolution completion rate and the estimated remaining equilibrium time, thus achieving accurate prediction of the leaching process. Combined with the phased monitoring strategy of the pH and temperature dual-parameter collaborative monitoring system, the temperature and time parameters can be dynamically adjusted according to the prediction results, ensuring that the leaching process is always in the optimal state and effectively avoiding the problems of incomplete dissolution or over-leaching. Attached Figure Description

[0050] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0051] Figure 1 This is a flowchart of the method of the present invention;

[0052] Figure 2 This is a diagram illustrating the technical implementation of the present invention;

[0053] Figure 3 This is a flowchart of the optimization algorithm of the present invention;

[0054] Figure 4 This is a flowchart of the model processing of the present invention. Detailed Implementation

[0055] The technical solutions of the present invention will be clearly and completely described below with reference to the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0056] like Figure 1 As shown, this invention provides a model prediction-optimized process for making Cornus officinalis health-preserving wine, comprising:

[0057] S1. The active ingredients of five medicinal materials, namely Cornus officinalis, mulberry, Poria cocos, jujube and ginseng, are tested using the corresponding standard methods. A cell wall degradation index evaluation system is established through mechanical strength testing. The cell wall degradation index of each medicinal material is measured and the comprehensive cell wall degradation index is calculated by preset ratio weights. A comprehensive quality evaluation function is established based on the content of active ingredients and the comprehensive cell wall degradation index. The raw materials are graded and coded as raw material quality grade values.

[0058] S2. Based on the raw material quality grade value and the comprehensive cell wall degradation index, a first-order dissolution kinetic model based on the equivalent active ingredient concentration is established. The modified Arrhenius equation is used to describe the relationship between the dissolution rate constant and the temperature and raw material properties. Based on this dissolution kinetic model, the single-objective parameter optimization is carried out in combination with the golden section search algorithm to obtain the corresponding optimal immersion temperature and optimal immersion time.

[0059] S3. Weigh the medicinal materials and mix them with the base liquor according to the preset ratio. After drying and pulverizing the medicinal materials, put them into the soaking container and start soaking according to the optimal soaking temperature and optimal soaking time.

[0060] S4. Install pH and temperature sensors in the immersion container to establish a dual-parameter collaborative monitoring system, and adopt a phased monitoring strategy to collect monitoring parameters in real time during the immersion process.

[0061] S5. Construct a random forest model, using the rate of change of monitoring parameters, immersion time, raw material quality grade, comprehensive cell wall degradation index, and pH-temperature synergistic index as input features. Simultaneously predict the dissolution completion rate and the estimated remaining equilibrium time. Based on the prediction results, dynamically adjust the temperature and time parameters in the immersion container. When the predicted dissolution completion rate reaches 95% and the estimated remaining equilibrium time is ≤1 day, end the dynamic adjustment process, and continue to wait for the predicted remaining equilibrium time to complete the immersion to obtain the immersion solution.

[0062] S6. Remove the maceration solution from the maceration container, separate and purify it using a two-stage filtration process, and conduct quality testing and verification on the product to obtain the Cornus officinalis health wine product.

[0063] like Figure 2 As shown, the technical approach of this invention is to construct a complete closed-loop control system, from precise evaluation of raw material quality to scientific setting of process parameters and intelligent control of the maceration process. First, the active ingredient content of the five medicinal herbs is detected using appropriate standard testing methods, and a cell wall degradation index evaluation system is established, quantifying the chemical quality and physical properties of the raw materials into numerical variables. Then, based on a first-order dissolution kinetic model of equivalent active ingredient concentration combined with a modified Arrhenius equation, the relationship between temperature, raw material quality, and dissolution rate is described. A golden section search algorithm is used for single-objective parameter optimization to determine the optimal temperature and time combination that achieves 95% dissolution completion, and a parameter fast lookup table is established to improve efficiency. During the maceration process, a random forest model is constructed using a phased dual-task learning framework. A dual-parameter collaborative monitoring system for pH and temperature is used to collect maceration parameters in real time, while simultaneously predicting the dissolution completion rate and the estimated remaining equilibrium time. Temperature and time parameters are dynamically controlled based on the prediction results. Finally, separation and purification are achieved through a two-stage filtration process. A quality consistency evaluation system and a batch file management system are established, realizing an intelligent process upgrade from traditional experience-driven to data-driven, ensuring the stability and consistency of the quality of the Cornus officinalis health wine product.

[0064] Specifically, in one embodiment of the present invention, step S1 includes:

[0065] First, the key active ingredients of the five medicinal herbs were precisely quantified using appropriate standard analytical methods. Specific tests included: iridoid glycoside content (calculated as monoglycosides, HPLC method) and organic acid content (HPLC method) of Cornus officinalis; anthocyanin content (HPLC method) and total sugar content (colorimetric method) of Morus alba; polysaccharide content (phenol-sulfuric acid colorimetric method) of Poria cocos; total sugar content (anthrone colorimetric method) of Jujube; and saponin content (HPLC method) of Ginseng. The required accuracy was ±0.05% for iridoid glycosides and ±0.1% for sugars.

[0066] Based on the completion of routine quality indicator tests, the degree of cell wall degradation of the medicinal materials was further evaluated through microscopic observation and mechanical strength testing. A cell wall degradation index D was established. cell The degree of degradation of the cell wall structure integrity of a single medicinal herb is quantified by measuring the crushing resistance per unit weight under standard pressure. cell A higher value indicates a more severe degree of cell wall deterioration, and a more easily soluble active ingredient.

[0067] Cell wall degradation index D cell The quantification employed a standardized mechanical strength testing method. First, the medicinal material samples were dried at a constant temperature of 60℃ until the moisture content was ≤5%. Then, they were pre-treated using a 20-mesh standard sieve to ensure relatively uniform particle size. During testing, a universal testing machine (such as the Instron 5969 model) was used, equipped with a 10mm diameter circular indenter. Standard test conditions were set as follows: compression speed 1mm / min, maximum pressure 50N, ambient temperature 25±2℃, and relative humidity 50±5%.

[0068] The specific operating procedure is as follows: Accurately weigh 1.000±0.005g of medicinal material sample and place it on the standard test platform. Start the compression program until the sample is completely broken, and record the maximum compressive force F when complete breakage is achieved. measured (Unit: N)

[0069] Establish a cell wall degradation index evaluation system. Due to the significant differences in cell wall structure among the five medicinal herbs, it is necessary to determine the cell wall degradation index D for each herb separately. cell,j (j = 1, 2, 3, 4, 5 correspond to Cornus officinalis, mulberry, Poria cocos, jujube, and ginseng, respectively), and then the comprehensive cell wall degradation index is calculated by weighting the proportions.

[0070] The cell wall degradation index of each medicinal material is calculated using the following formula:

[0071]

[0072] Where F base,j The reference breaking force for each medicinal material is the standard mechanical strength of the cell wall under ideal, intact conditions, serving as a reference benchmark for evaluating the degree of deterioration. F measured,jTo measure the actual breaking force, the maximum compressive force required to break the medicinal material under standard testing conditions is measured, reflecting the actual strength of the cell wall. The baseline breaking force is the standard breaking force for the corresponding medicinal material, representing the standard mechanical strength of the cell wall in an ideal, intact state, serving as a reference for evaluating the degree of deterioration. The measured breaking force is the maximum compressive force required to break the medicinal material under standard testing conditions, reflecting the actual strength of the cell wall. The baseline values ​​for each medicinal material are determined through standard tests. Fresh, intact, undamaged samples with moisture content conforming to the relevant pharmacopoeia regulations are selected. A unified physical and mechanical testing method is used, and the test is repeated more than 30 times under the same testing conditions. The average value is taken as the baseline value. The baseline values ​​for each medicinal material are as follows: Cornus officinalis F... base,1 =28.5N, mulberry F base,2 =22.3N, Poria F base,3 =35.8N, Jujube F base,4 =31.2N, Ginseng F base,5 =26.9N.

[0073] The overall cell wall degradation index was calculated using a ratio-weighted average method.

[0074]

[0075] Where w j The weighting coefficients for each medicinal material are determined based on the feed ratio of 5:5:4:5:1: w 山茱萸 =w 桑椹 =w 大枣 =0.25, w 茯苓 =0.2, w 人参 =0.05.

[0076] When the cell walls of medicinal materials remain intact, considerable pressure is required to break them. measured,j Approaching F base,j D cell,j Approaching 0; when the cell wall deteriorates, the pressure required for breakage decreases, F measured,j Significantly smaller than F base,j D cell,j The value increases.

[0077] Establish grading standards for the degree of deterioration of various medicinal materials: D cell,j ≤0.15 indicates slight degradation, 0.15 <D cell,j ≤0.35 indicates moderate degradation, 0.35 <D cell,j ≤0.55 indicates severe degradation, D cell,j A value >0.55 indicates severe degradation.

[0078] Based on the degree of deterioration of each medicinal material, a comprehensive cell wall deterioration index is calculated. The grading criteria for the composite index have been adjusted accordingly: To account for minor degradation, To be considered as moderately degraded, For comprehensive severe degradation, The overall situation has deteriorated significantly.

[0079] A comprehensive quality evaluation function based on dissolution kinetics weights was established according to the content of active ingredients and the cell wall degradation index. A nonlinear model considering diffusion resistance was used for the quality evaluation of each medicinal material.

[0080]

[0081] Among them, Q total The overall quality score of the medicinal materials; C i The content of the i-th active ingredient includes the monoglycoside and organic acid content of Cornus officinalis, the anthocyanin and total sugar content of Morus alba, the polysaccharide content of Poria cocos, the total sugar content of Jujube, and the saponin content of Ginseng. This is the absolute concentration of the key active ingredients in the medicinal materials, which directly determines the efficacy and value of the product; α i The dissolution difficulty index of active ingredients (cornucigenin α1 = 1.2, organic acid α2 = 0.8, ginsenoside α3 = 1.5, polysaccharide α4 = 1.0, anthocyanin α5 = 0.9) characterizes the relative ease with which different types of active ingredients are transferred from the cell to the solvent, reflecting differences in molecular structure and cell wall affinity; W i R is the importance weighting coefficient, representing the trade-off coefficient between the contribution of each active ingredient to the overall efficacy and quality of the product; barrier,i is the cell wall diffusion resistance constant of the i-th active ingredient, which characterizes the resistance encountered by the active ingredient when it undergoes mass transfer through the cell wall and is related to molecular size and cell wall pore structure; β = 0.15 is the cell wall deterioration dissolution promotion coefficient, which is a quantitative parameter characterizing the effect of cell wall deterioration on the reduction of diffusion resistance and reflects the correlation between cell wall integrity and dissolution capacity. This is a comprehensive cell wall degradation index.

[0082] The physical meaning of this evaluation function is that the contribution of the active ingredient content to the overall quality is constrained by the diffusion resistance of the cell wall. Moderate degradation of the cell wall can reduce the diffusion resistance and thus improve the dissolution efficiency.

[0083] Based on Q total The raw materials are divided into three grades: Grade A (superior) Q total ≥85, Grade B (First Class) 65≤Q total <85, Grade C (Qualified) 45≤Q total <65. After each batch of raw materials is graded, the quality grade is coded as a numerical variable G. raw(Grade A = 3, Grade B = 2, Grade C = 1).

[0084] Specifically, in one embodiment of the present invention, step S2 includes:

[0085] Based on the raw material quality grade G obtained in step S1 raw and cell wall degradation index D cell A simplified first-order kinetic model was established, and the initial immersion temperature and time parameters were determined through a single-objective optimization algorithm.

[0086] A first-order dissolution kinetic model based on comprehensive active ingredients was established. Considering that the five medicinal herbs contain different types of active ingredients and have different dissolution characteristics, a unified kinetic model was established using the concept of equivalent active ingredient concentration.

[0087] The dissolution process of the comprehensive active ingredients can be simplified into a first-order kinetic equation:

[0088] C(t)=C max ·[1-exp(-k·t)]

[0089] Where C(t) is the equivalent active ingredient concentration (mg / L) at time t, a comprehensive concentration index that standardizes the concentration of all active ingredients of the five medicinal herbs according to their contribution to product quality, and can comprehensively reflect the overall quality level of the extract; C max The theoretical maximum equivalent dissolution concentration (mg / L) represents the maximum concentration of active ingredient that can be dissolved in an infinitely long time, signifying the theoretical dissolution limit of the raw material, which is constrained by the quality of the raw material and the integrity of the cell wall; k is the overall dissolution rate constant (day). -1 ), the rate parameter of the transfer of active ingredients from the solid phase to the liquid phase. The larger the value, the faster the dissolution. It is affected by factors such as temperature, cell wall state and raw material quality; t is the immersion time (days), which is the actual time elapsed from the start of immersion to the current moment.

[0090] The relationship between the overall dissolution rate constant k and temperature and overall feed properties follows the modified Arrhenius equation:

[0091]

[0092] Where k0 = 0.025day -1 E is the reference rate constant. a=28000 J / mol is the apparent activation energy; R = 8.314 J / (mol·K) is the gas constant; T is the absolute temperature (K); γ = 0.15 is the cell wall degradation promoting coefficient, which characterizes the promoting effect of cell wall degradation on dissolution rate and reflects the relationship between cell wall integrity and mass transfer resistance; μ = 0.12 is the quality grade correction coefficient, which characterizes the correction coefficient of raw material quality grade on dissolution kinetics. The lower the quality, the stronger the driving force required to achieve full dissolution.

[0093] The physical meaning of this model is that increased temperature can increase the molecular diffusion rate, the comprehensive cell wall degradation index reflects the average mass transfer resistance level of all medicinal materials, and lower-quality raw materials require stronger driving forces to achieve full dissolution. The weighted averaging method considers the differences among medicinal materials while maintaining the model's simplicity and practicality. (3-G) raw The value represents the difference between the quality grade and the highest grade. The larger the value, the lower the quality and the stronger the correction effect on the dissolution rate.

[0094] The golden section search algorithm is used for parameter optimization. The golden section method is a classic one-dimensional search algorithm that finds the optimal solution by progressively narrowing the search interval. Compared to complex swarm intelligence algorithms, the golden section method is computationally simple, convergently stable, and easy to implement.

[0095] like Figure 3 As shown, the specific optimization process of the golden section search algorithm includes:

[0096] Step 1: Set search parameters based on the raw material quality grade value G. raw and comprehensive cell wall degradation index Set temperature search range [T] min ,T max ] and time search range [t min ,t max ], set precision requirements ∈ T and ∈ t The specific configuration strategy is as follows:

[0097] For Grade A raw materials (G) raw =3), due to the relatively intact cell wall structure, the temperature range is set to [313K, 323K], and the time range is [25 days, 35 days]; for Grade B raw materials (G raw =2), set the temperature range [318K, 328K], and the time range [28 days, 38 days]; for Grade C raw materials (G raw =1), due to the need for a stronger dissolution driving force, the temperature range [323K, 333K] and the time range [30 days, 40 days] are set. Simultaneously, the temperature range is fine-tuned based on the comprehensive cell wall degradation index: when When the above temperature range drops by 2 - 3K each, because the dissolution resistance decreases when the degree of cell wall deterioration is high. The precision requirement is set to ∈ T = 0.1K (temperature), ∈ t = 0.5 days (time).

[0098] Step 2: Construct the optimization objective function:

[0099]

[0100] The goal is to find the optimal temperature-time combination that makes the dissolution completion rate reach 95%; where C(t) is calculated by the first-order dissolution kinetic model, and the comprehensive dissolution rate constant k is determined by the modified Arrhenius equation.

[0101] Step 3: Calculate the initial objective function value. Select two golden section points T1 = T min + 0.382×(T max - T min ) and T2 = T min + 0.618×(T max - T min ) within the temperature search range. Here, 0.382 and 0.618 are the standard division ratios of the golden section method, determined based on mathematical optimization theory, which can ensure the highest search efficiency. Calculate the objective function values F(T1, t) and F(t2, t) corresponding to these two division points respectively, where the time parameter t can be set as the midpoint value of the search range in the initial stage.

[0102] Step 4: Conduct optimization iteration in the temperature dimension. When F(T1, t) < F(T2, t), it means the optimal solution is within the interval [T min , T2], so update the search range to [T min , T2] and re-select the golden section points. When F(T1, t) ≥ F(T2, t), the optimal solution is within the interval [T1, T max , then update the search range to [T1, T max and re-select the golden section points; repeat this process until the length of the search range is less than the precision requirement ∈ T .

[0103] Step 5: Conduct optimization iteration in the time dimension. Fix the optimal temperature obtained in Step 4 Within the time search range [t min , t max , use the same golden section method to find the optimal time parameter. Calculate the division points t1 = t min + 0.382×(t max - t min ) and t2 = t min + 0.618×(tmax -t min Compare the objective function values. and Update the search time range based on the comparison results.

[0104] Step Six: Alternately execute Steps Four and Five until the changes in temperature and time parameters in two consecutive iterations are less than the accuracy requirement (|T). k+1 -T k |<∈ T And |t k+1 -t k |<∈ t When k represents the iteration number (or the maximum number of iterations is reached), the optimization process ends, and the optimal immersion temperature T is output. opt and optimal immersion time t opt .

[0105] Specifically, in one embodiment of the present invention, the method further includes:

[0106] Establish a parameter quick lookup table to classify raw material quality grade G raw m1 level and comprehensive cell wall degradation index The preset m2 intervals are combined to form m1×m2 combinations of raw material properties; specifically set as: G raw There are 3 levels (Level A = 3, Level B = 2, Level C = 1). Seven intervals are set (divided by 0.1 increments: [0-0.1), [0.1-0.2), [0.2-0.3), [0.3-0.4), [0.4-0.5), [0.5-0.6), [0.6-0.7]), resulting in m1×m2=3×7=21 possible combinations of raw material properties. In this embodiment, in the division... When considering the range, 0.7 is used as the upper limit because 0.7 is far beyond the critical value of 0.55 for "severe degradation". This is not its theoretical maximum value. The theoretical maximum value should be close to 1.0. This is the practical upper limit in engineering applications. Raw materials exceeding this value are not recommended for production.

[0107] The optimal immersion temperature T for each combination is pre-calculated using the golden section search algorithm. opt and optimal immersion time t opt A parameter lookup table is created and stored; this process is completed during system setup to ensure that each combination has corresponding optimization parameters. The calculation results are stored in tabular form, with rows representing quality levels and columns representing cell wall degradation index ranges. Table elements are (T... opt ,t opt ) parameter pairs.

[0108] In production, the corresponding process parameters are directly queried based on the raw material quality grade value and the comprehensive cell wall degradation index in step S1; for raw material property parameters that are between the middle values ​​of the preset range, the corresponding temperature and time parameters are calculated using the bilinear interpolation method.

[0109] Specifically, in actual production, the raw material quality grade value G obtained in step S1 is used... raw and comprehensive cell wall degradation index First, determine Which preset range does it fall into? If If it is exactly equal to the interval boundary value, then directly query the parameter lookup table to obtain the corresponding (T) value. opt ,t opt This direct query method avoids redundant optimization calculations and greatly improves the efficiency of parameter setting.

[0110] For raw material properties that fall within the middle of a preset range, bilinear interpolation is used to calculate the corresponding temperature and time parameters. The specific process is as follows: assuming the actual measured parameter combination is... Find the four nearest neighbor preset grid points And its corresponding parameter values, and then through the bilinear interpolation formula:

[0111]

[0112] Where T ij Represents grid points The corresponding optimal temperature. Time parameter t interpolated The same interpolation formula is used for calculation.

[0113] For raw material properties falling within the middle of a preset range, bilinear interpolation is used to calculate the corresponding temperature and time parameters. This method is suitable for normal operating conditions where the raw material quality grade is between 1 and 3, and the comprehensive cell wall degradation index is between 0 and 0.7. When raw material parameters approach boundary values, the interpolation method remains effective, but boundary constraint checks are required. The specific process is as follows: Assuming the actual measured parameter combination is the raw material quality grade and the comprehensive cell wall degradation index, find the four nearest neighbor preset grid points and their corresponding parameter values, then calculate the temperature parameter using the bilinear interpolation formula. The time parameter is calculated using the same interpolation formula. Whether the parameters are obtained through direct lookup or interpolation, an engineering rationality check is required: the temperature parameter should be within the range of [313K, 333K], and the time parameter should be within the range of [20 days, 45 days]. If the interpolation result exceeds the reasonable range, boundary value constraints are applied. Simultaneously, verify whether the theoretical dissolution completion rate can reach the target requirement of 95% to ensure the effectiveness of the parameter settings.

[0114] This fast parameter lookup method combines pre-computation and bilinear interpolation, ensuring both high efficiency in parameter acquisition and continuity and smoothness in parameter changes.

[0115] Specifically, in one embodiment of the present invention, step S3 includes:

[0116] Strictly according to the weight ratio m 山茱萸 :m 桑椹 :m 茯苓 :m 大枣 :m 人参 Weigh each herb in a ratio of 5:5:4:5:1, ensuring the total weight is within the range of M. total =200g±2g, mixed with 53-degree Maotai-flavor liquor at a mass ratio of 1:10, i.e., the volume of the liquor is V 酒 =2000mL±20mL.

[0117] Before feeding, all medicinal materials undergo standardized pretreatment, including drying at a constant temperature of 60℃ until the moisture content w ≤ 10%, followed by pulverization using a 20-40 mesh sieve to ensure the particle diameter is within the range of 0.42-0.84 mm. This particle size range is determined by balancing the mass transfer area and the difficulty of subsequent separation.

[0118] After pretreatment, the actual particle size distribution of each batch of medicinal materials was measured, and the specific surface area S was calculated. specific This is used for corrective calculations of mass transfer rate in subsequent process monitoring. Specific surface area is determined using the BET method, with a typical value of Sa. specific =0.15-0.25m 2 / g.

[0119] Use food-grade stainless steel immersion containers with a volume of at least 3L. The inner surface of the container must be smooth and free of dead corners for easy cleaning and disinfection. The immersion container must be equipped with a sealed lid to reduce alcohol evaporation and external contamination.

[0120] The pre-treated medicinal materials are added into the soaking container in sequence according to their weight ratio, using a layered feeding method: the bottom layer is filled with harder medicinal materials (Poria cocos and Cornus officinalis), the middle layer is filled with medium-hard medicinal materials (jujube), the top layer is filled with softer medicinal materials (mulberry), and the top layer is filled with ginseng.

[0121] Based on the optimal immersion temperature T obtained in step S2 opt and optimal immersion time t opt Begin the soaking process. The soaking time is defined as the moment when the liquor and medicinal herbs are in complete contact. After adding the raw materials, slowly add the pre-treated 53-degree Maotai-flavor liquor at a rate of 100-150 mL / min to avoid uneven suspension of the medicinal herbs due to impact. After adding the liquor, gently stir with a clean stainless steel stirring rod for 2-3 minutes to ensure full contact between the medicinal herbs and the liquor.

[0122] The immersion environment temperature is controlled at T opt Within a range of ±1K, the relative humidity should be controlled at 50-70%. The immersion container should be placed in a dark environment to prevent UV damage to the active ingredients. The immersion area should be kept clean and disinfected regularly.

[0123] Specifically, this invention establishes a dual-parameter collaborative monitoring system by deploying pH and temperature sensors within the immersion container. The sensor deployment must adhere to the following rules:

[0124] In the food-grade stainless steel immersion container, the pH sensors are arranged in a three-point configuration: the main sensor is located at the center of the container, one-third of the way from the bottom; the auxiliary sensor is located at the edge of the container, 5 cm from the container wall and half the way from the bottom, to monitor pH changes in the boundary layer; and the calibration sensor is located at the top of the container, 2 cm below the liquid surface, to monitor pH fluctuations near the liquid surface. All three pH sensors have a measurement accuracy of ±0.01 and a measurement range of pH 2.0-7.0, specifically calibrated for acidic immersion environments.

[0125] The temperature sensors are arranged in a layered configuration: the bottom layer temperature sensor is located 5mm from the bottom of the container, monitoring temperature changes near the heating element; the middle layer temperature sensor is located in the center of the herbal layer, at 1 / 3 of the container's height from the bottom, monitoring the main temperature of the infusion solution; the top layer temperature sensor is located 1cm below the liquid surface, monitoring heat loss and temperature distribution uniformity. The temperature sensors are Pt100 platinum resistance thermometers, with a measurement accuracy of ±0.1K, a measurement range of 273K-353K, and a response time ≤3 seconds.

[0126] Establish a phased monitoring strategy to adapt to the different characteristics of the immersion process:

[0127] Early rapid dissolution period (days 1-7):

[0128] The key focus is on monitoring pH changes; the normal pH range should be 3.8 ≤ pH ≤ 4.2. During this stage, the cell walls of the medicinal materials begin to rupture, rapidly releasing small molecule active ingredients such as organic acids, leading to a significant drop in pH. Monitoring the rate of pH change is crucial. (where Δt = 1 day), to determine whether the dissolution rate is normal.

[0129] Mid-term equilibrium transition period (days 8-21):

[0130] The focus shifts to monitoring temperature stability, with temperature fluctuations controlled within ΔT ≤ 1K. During this stage, large molecular active ingredients such as polysaccharides and saponins begin to dissolve, and the mass transfer process is primarily affected by temperature. Temperature stability indices are calculated. (where std(T) is the standard deviation of temperature within the sliding window), to evaluate the temperature control effect.

[0131] Late equilibrium dissolution period (days 22-35):

[0132] A comprehensive coordination assessment of the two parameters was conducted by calculating the pH-temperature synergy index S. synergy =corr(pH,T) (Pearson correlation coefficient) is used to determine the maceration process. Under normal circumstances, S... synergy A value between -0.3 and 0.3 should indicate a reasonable correlation between pH and temperature changes. When |S synergy When |>0.5, it indicates that there may be abnormal coupling in the system, and the process parameters need to be fine-tuned.

[0133] Different sampling frequencies are set according to the monitoring focus at different stages: In the early rapid dissolution period, pH data is collected every 30 minutes, and temperature data every 5 minutes; in the middle transition period, pH data is collected every hour, and temperature data every 5 minutes; in the later equilibrium period, both parameters are collected every hour. The sampling frequency is determined based on the leaching kinetics: in the early rapid dissolution period, drastic pH changes require high-frequency monitoring, and temperature fluctuations significantly affect mass transfer, requiring intensive sampling; in the middle and later stages, as changes slow down, the frequency can be appropriately reduced. In practical applications, the sampling frequency can be adjusted within ±50% according to the specific equipment accuracy and cost requirements, the key being to ensure timely capture of abnormal fluctuations. A data smoothing filtering algorithm is established, using a 5-point moving average to preprocess the raw data to eliminate the influence of sensor noise and short-term fluctuations. Among them, pH filtered (t) represents the pH value after filtering at time t; pH raw (t+i) represents the original pH measurement value at time (t+i); i is the time offset, with a value range of i∈{-2,-1,0,+1,+2}, representing the i sampling points before / after the current time; at the same time, a data anomaly detection mechanism is established, and when a single measurement value deviates from the moving average by more than 3 standard deviations, it is marked as abnormal data and corrected by interpolation method.

[0134] It should be noted that in this embodiment, the original pH and temperature measurements are obtained by fusing data from multiple sensors of the same type into a single representative value. Specifically, a weighted summation method is used, and then the fused data is filtered to finally obtain a single pH and temperature value.

[0135] Specifically, in one embodiment of the present invention, step S5 includes:

[0136] A random forest model based on a phased dual-task learning framework is constructed to simultaneously predict two key objectives: dissolution completion E. completion∈[0,1] and the expected remaining equilibrium time t remaining (sky).

[0137] The model training process is as follows:

[0138] First, historical data from at least 100 complete immersion batches were collected as the training set. This historical data includes complete monitoring records for each batch from day 1 to day 35, along with corresponding label data such as final dissolution completion and actual immersion time. The training data was divided using an overlapping time window strategy: the first sub-model was trained using data from days 1-15, focusing on learning the characteristic patterns of the rapid dissolution period; the second sub-model was trained using data from days 8-35, focusing on learning the characteristic patterns of the balanced dissolution period. This 8-15 day overlapping training period ensured that both sub-models could effectively identify and process input features during the transition phase, providing a reliable predictive foundation for subsequent dynamic weight fusion. Both sub-models used the same architecture and multi-task loss function, but due to the different emphasis on the training data's time window, the final learned decision rules and feature weights differed.

[0139] The first sub-model focuses on rapidly changing features such as pH rate of change, with 50 decision trees, a maximum depth of 6 layers, and a feature sampling ratio of 70%. This relatively simple structure can quickly capture the drastic changes in the early stages of leaching. The second sub-model needs to handle more complex and subtle feature relationships, therefore it has 80 decision trees, a maximum depth of 8 layers, and a feature sampling ratio of 80%. The more complex structure helps to identify subtle changes during the equilibrium period. Both sub-models are trained using a multi-task loss function.

[0140] L total =λ1·L completion +λ2·L remaining

[0141] Where L completion L represents the mean square error of dissolution completion. remaining The Huber loss function is for the remaining time, with weight coefficients λ1 = 0.7 and λ2 = 0.3. The Huber loss function is defined as:

[0142]

[0143] Where ∈ = 1.5 days is the tolerance parameter.

[0144] The training process uses 5-fold cross-validation to evaluate model performance. The model training is considered to have converged when the validation set loss no longer decreases after 10 consecutive epochs and the mean absolute errors of the two prediction tasks are less than 0.05 and 1.5 days, respectively.

[0145] like Figure 4 As shown, in the actual prediction process, real-time pH and temperature data are first obtained from the monitoring system in step S4. These data are then processed through multi-sensor fusion and a 5-point moving average filter to form representative pH and temperature values ​​for the current moment. Simultaneously, the comprehensive cell wall degradation index calculated in step S1 is used. and raw material quality grade value G raw (Grade A = 3, Grade B = 2, Grade C = 1).

[0146] Based on this fundamental data, the components of the six-dimensional feature vector are calculated in real time: First, the rate of pH change within the sliding window. Where Δt = 1 day; second, temperature stability index Where std(T) is the standard deviation of temperature within the sliding window; third, the immersion time t current (Actual number of days from the start of soaking to the present); Fourth, raw material quality grade code G raw Fifth, the comprehensive cell wall degradation index. Sixth, pH-temperature synergistic index S synergy =corr(pH,T), which is the Pearson correlation coefficient between pH and temperature.

[0147] During prediction execution, both the first and second sub-models are invoked simultaneously, using the same six-dimensional feature vector as input, to obtain two prediction results Y1 = [E]. completion,1 ,t remaining,1 ] and Y2=[E completion,2 ,t remaining,2 ].

[0148] Then, based on the current soaking time t current Calculate the dynamic weighting coefficients:

[0149]

[0150] w1 represents the weights of the first sub-model, w2 represents the weights of the second sub-model, and the final prediction result Y pred Obtained through dynamic weight fusion:

[0151] Y pred =w1·Y1+w2·Y2

[0152] This dynamic weighting mechanism allows the system to rely primarily on the rapid feature recognition capabilities of the early sub-models in the early stages of immersion, and primarily on the equilibrium feature analysis capabilities of the mid-to-late stages of immersion. The weighting transformation process is smooth and continuous, avoiding abrupt changes in the prediction results.

[0153] After obtaining the model prediction results, the process of dynamically controlling the temperature and time parameters in the immersion container based on the prediction results includes:

[0154] When the random forest model predicts abnormal dissolution progress, corresponding control measures are initiated: when the dissolution completion rate is lower than expected (E... completion <0.7 and t current If the soaking time is >21 days, extend the soaking time by 3-5 days, the specific extension time Δt extend =5-2×E completion When the rate of pH change is abnormal (|ΔpH) rate When |>0.03 / day), adjust the temperature ΔT=-2K×sign(ΔpH) rate To stabilize chemical equilibrium; when the predicted remaining time is too long (t) remaining When the temperature is stable (>8 days) and the temperature is increased by ΔT = 1-2K, the soaking time should be appropriately extended; when the predicted remaining time is too short (t remaining If the leaching time is less than 2 days and the dissolution rate is insufficient, extend the leaching time and closely monitor the process to prevent over-leaching.

[0155] Control termination condition: When the predicted dissolution completion rate reaches E completion ≥0.95 and estimated remaining equilibrium time t remaining When the time is ≤1 day, the dynamic control process ends, and the process continues to wait for the predicted remaining equilibrium time to complete the immersion. At this point, no further artificial intervention in temperature or time is performed, allowing the immersion process to naturally reach chemical equilibrium.

[0156] Specifically, in one embodiment of the present invention, step S6 includes:

[0157] Remove the immersion solution completely from the food-grade stainless steel immersion container to avoid quality changes due to prolonged contact time.

[0158] The two-stage filtration process employs a step-by-step refining strategy. The first-stage coarse filtration uses a 200-mesh stainless steel filter with a pore size of 0.074mm, primarily removing incompletely dissolved medicinal material particles and large molecular impurities. The filtration rate is strictly controlled at 50-80mL / min, and the filter material is selected from 316L food-grade stainless steel. After the first-stage coarse filtration, most suspended particles in the extract are removed, but small colloidal particles and particulate impurities may still be present.

[0159] The secondary fine filtration uses a 0.45μm ceramic membrane filter. This pore size effectively removes bacteria, yeast, and other microorganisms while preserving the integrity of active ingredients. During filtration, the filter membrane pressure differential needs to be monitored regularly. When the pressure differential exceeds 0.2MPa, the filter membrane should be cleaned or replaced promptly to ensure stable filtration performance. The product after secondary fine filtration should have a clear and transparent appearance, with a turbidity value controlled below 2 NTU.

[0160] Quality testing and verification employs a multi-dimensional indicator system to comprehensively evaluate the sensory quality and active ingredient content of the product. Color (a*) value testing uses a colorimeter, with a normal range of 8.5-12.5, reflecting the depth of the red color. An excessively high a value indicates potential oxidation issues, while an excessively low a value may indicate insufficient dissolution of the active ingredients. Clarity (NTU) value testing uses a turbidimeter, with a passing standard of ≤2.0 NTU, ensuring good transparency and visual appeal. Dry matter content testing uses a gravimetric method, drying a certain volume of sample at 105℃ to constant weight and calculating the dry matter content. The normal range is 15-25 g / L, reflecting the total amount of dissolved solids in the leaching solution.

[0161] The content of key active ingredients was determined using corresponding standard methods, specifically: mononosides in Cornus officinalis (HPLC), anthocyanins in mulberry (HPLC), polysaccharides in Poria cocos (phenol-sulfuric acid colorimetric method), total sugars in jujube (anthrone colorimetric method), and saponins in ginseng (HPLC). The content standards for each active ingredient were based on the raw material quality grade value G in step S1. raw Dynamic adjustments will be made: Products made from Grade A raw materials should have 95%-105% of the standard value for each active ingredient; products made from Grade B raw materials should have 85%-95%; and products made from Grade C raw materials should have 75%-85%. Sensory evaluation will be conducted blindly by a team of professional sommeliers, scoring based on four dimensions: color, aroma, taste, and finish. The maximum score is 100 points, and a passing score is ≥80 points.

[0162] A quality consistency evaluation system is established, and process stability is assessed by calculating the batch-to-batch coefficient of variation for each testing indicator. The formula for calculating quality consistency evaluation is:

[0163]

[0164] Q u This represents the overall quality score for batch u. For the average quality score of U batches, the target control CV is... consistency <5%.

[0165] Establish a complete batch record management system to record information across the entire chain, from raw material quality grade, cell wall degradation index, process parameter settings, process monitoring data, random forest model prediction results to final test data, forming a complete quality traceability system. Each batch is assigned a unique traceability code, enabling rapid identification of the cause and implementation of corrective measures when quality issues arise.

[0166] A continuous optimization mechanism for the model is established. After each batch is completed, data such as the actual dissolution completion rate and the actual soaking time are added to the training set. The parameters of the random forest model are updated through incremental learning, so that the model prediction accuracy is continuously improved with the accumulation of data, and finally, a Cornus officinalis health wine product that meets the quality requirements is obtained.

[0167] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A model prediction optimization-based Cornus officinalis health wine extraction process, characterized in that, The method comprises the following steps: S1, active ingredient detection is performed on each of the five medicinal materials, i.e., cornus officinalis, mulberry, poria cocos, jujube and ginseng, by using a corresponding standard method, a cell wall deterioration index evaluation system is established by mechanical strength test, the cell wall deterioration index of each medicinal material is determined, and a comprehensive cell wall deterioration index is calculated by using a preset proportion weight; a comprehensive quality evaluation function is established based on the active ingredient content and the comprehensive cell wall deterioration index, the raw materials are classified and coded into raw material quality grade numerical values; S2, a first-order dissolution kinetics model based on the equivalent active ingredient concentration is established according to the raw material quality grade numerical values and the comprehensive cell wall deterioration index, a modified Arrhenius equation is used to describe the relationship between the dissolution rate constant and the temperature and the properties of the raw materials, a single-objective parameter optimization is performed based on the dissolution kinetics model and a golden section search algorithm, and the corresponding optimal soaking temperature and optimal soaking time are obtained; S3, the medicinal materials and base liquor are weighed according to a preset proportion, the medicinal materials are dried and crushed for pretreatment, and then are put into a soaking container, and soaking is started according to the optimal soaking temperature and the optimal soaking time; S4, a pH sensor and a temperature sensor are arranged in the soaking container to establish a double-parameter cooperative monitoring system, and a staged monitoring strategy is used to collect the monitoring parameters in the soaking process in real time; S5, a random forest model is constructed, the change rate of the monitoring parameters, the soaking time, the raw material quality grade numerical value, the comprehensive cell wall deterioration index and a pH-temperature cooperative index are used as input features, the dissolution completion degree and the predicted remaining balance time are predicted, the temperature and the time parameters in the soaking container are dynamically regulated according to the prediction results, the dynamic regulation process is ended when the predicted dissolution completion degree reaches 95% and the predicted remaining balance time is less than or equal to 1 day, and the soaking is continued until the predicted remaining balance time is completed to obtain the soaking liquid; S6, the soaking liquid is taken out of the soaking container, a two-stage filtration process is used for separation and purification, and quality detection and verification are performed on the product to obtain the cornus officinalis health wine product.

2. The model prediction optimization-based processing technology of Cornus officinalis wine according to claim 1, characterized in that, The calculation formula of the comprehensive quality evaluation function is as follows: Wherein, Q total is the comprehensive quality score of medicinal materials; C i is the content of the i th active ingredient, including the content of morroniside and organic acid of Cornus officinalis, the content of anthocyanin and total sugar of Morus alba, the content of polysaccharide of Poria cocos, the content of total sugar of Jujube, and the content of ginsenoside of Panax ginseng; α i is the dissolution difficulty index of active ingredients; W i is the importance weight coefficient; R barrier,i is the cell wall diffusion resistance constant of the i th active ingredient; β is the cell wall degradation dissolution promotion coefficient; is the comprehensive cell wall degradation index.

3. The model prediction optimization-based processing technology of Cornus officinalis wine according to claim 2, characterized in that, The cell wall deterioration index is quantified by a standardized mechanical strength test, and the calculation formula of the cell wall deterioration index of a single medicinal material is as follows: Wherein, j = 1, 2, 3, 4, 5 respectively corresponding to dogwood, mulberry, tuckahoe, jujube, ginseng; F base,j F is the reference crushing force corresponding to the medicinal material; measured,j F is the measured crushing force; The comprehensive cell wall deterioration index is calculated by using a weighted average: wherein w j is the weight weight coefficient of each medicinal material, determined according to the preset proportion.

4. The model prediction optimization-based cornus officinalis wine brewing process according to claim 1, wherein, The calculation formula of the first-order dissolution kinetics model is as follows: C(t) = C max • [1 - exp(-k · t)] where C(t) is the equivalent active ingredient concentration at time t; C max is the theoretical maximum equivalent dissolution concentration; k is the overall dissolution rate constant; t is the soaking time; The relationship between the comprehensive dissolution rate constant k and the temperature and the properties of the raw materials follows a modified Arrhenius equation: where k0 is a reference rate constant, E a is an apparent activation energy, R is a gas constant, T is an absolute temperature, γ is a cell wall deterioration promotion coefficient, μ is a quality grade correction coefficient, G raw is a raw material quality grade value.

5. The model prediction optimization-based processing technology of Cornus officinalis wine according to claim 4, characterized in that, The specific optimization process of the golden section search algorithm includes the following steps: Step 1, set search parameters, according to the raw material quality grade value G raw and the comprehensive cell wall degradation index Set temperature search range [T min , T max ] and time search range [t min , t max ], set precision requirements ∈ T and ∈ t ; Step 2, constructing an optimization objective function: The objective is to find the optimal temperature and time combination that makes the dissolution completion degree reach 95%; Step three, calculate the initial objective function value, select two golden section points T1 = T min + 0.382 x (T max -T min ) and T2 = T min + 0.618 x (T max -T min ) in the temperature search interval, respectively calculate the objective function values F(T1, t) and F(T2, t); Step four, if F(T1, t) < F(T2, t), then T = T1, otherwise T = T2; Step four, temperature dimension optimization iteration is carried out, when F(T1, t) < F(T2, t), then the search interval is updated as [T1, T2] and the golden section point is reselected, when F(T1, t) ≥ F(T2, t), then the search interval is updated as [T1, T2] and the golden section point is reselected; min max ;​ Step 5, time dimension optimization iteration, the optimal temperature obtained in step 4 is fixed, and the same golden section method is used to find the optimal time parameter in the time search interval; Step six, alternately execute step four and step five, when the temperature and time parameter changes of two consecutive iterations are both less than the precision requirement or the maximum number of iterations is reached, then end the optimization process, output the optimal soaking temperature T opt and the optimal soaking time t opt .

6. The model prediction optimization-based processing technology of Cornus officinalis wine according to claim 5, characterized in that, The method further comprises: establishing a parameter quick query table, combining m1 levels of the raw material quality grade G raw with the comprehensive cell wall degradation index to form m1*m2 kinds of raw material property combination cases; using a golden section search algorithm to pre-calculate the optimal steeping temperature T opt and the optimal steeping time t opt corresponding to each combination, establishing a parameter query table and storing; in production, directly querying the corresponding process parameters according to the raw material quality grade value and the comprehensive cell wall degradation index of step S1; for the case that the raw material property parameter is between the intermediate values of the preset intervals, using a bilinear interpolation method to calculate the corresponding temperature and time parameters.

7. The model prediction optimization-based processing technology of Cornus officinalis wine according to claim 1, characterized in that, The random forest model includes a first sub-model and a second sub-model; when prediction is performed, the two sub-models are called simultaneously to process the same input feature vector, and the prediction results of the dissolution completion degree and the remaining balance time are obtained, respectively, and the final prediction result is obtained by fusing the prediction results of the two sub-models based on the dynamic weight of the current soaking time.

8. The model prediction optimization-based processing technology of Cornus officinalis wine according to claim 1, characterized in that, In step S5, the strategy for dynamic regulation is as follows: When the dissolution process lags, the soaking time is extended; when the pH fluctuates abnormally, the soaking temperature is adjusted to stabilize the chemical equilibrium; when the predicted remaining time is abnormal, the soaking temperature and time are adjusted simultaneously; when the predicted dissolution completion degree reaches 95% and the predicted remaining balance time is ≤1 day, the dynamic regulation process is ended, and the soaking is continued until the predicted remaining balance time is completed.

9. The model prediction optimization-based processing technology of Cornus officinalis wine according to claim 1, characterized in that, The preset ratio in step S3 is 5:5:4:5:1 山茱萸 :m 桑椹 :m 茯苓 :m 大枣 :m 人参 The base liquor is 53-degree Maotai-flavor liquor, which is mixed with the medicinal materials at a mass ratio of 1:10; the pretreatment of the medicinal materials includes constant temperature drying at 60°C until the water content is less than or equal to 10% and powdering through a 20-40 mesh screen.

10. The model prediction optimization-based processing technology of Cornus officinalis wine according to claim 1, characterized in that, In step S6, the secondary filtration process includes: primary coarse filtration using a 200-mesh stainless steel filter screen to remove medicinal material particles and macromolecular impurities, with a filtration rate controlled at 50-80 mL / min; secondary fine filtration using a 0.45-μm ceramic membrane filter to ensure product clarity; Quality detection verification includes color a* value detection, clarity NTU value detection, dry matter content detection, key active ingredient content detection, and sensory evaluation; a quality consistency evaluation system is established, the process stability is evaluated by calculating the batch-to-batch variation coefficient of each detection index, and the variation coefficient control target is set to be less than 5%.

Images